code review 4185053: http: introduce Header type, implement with net/textproto

http: introduce Header type, implement with net/textproto
textproto: introduce Header type
websocket: use new interface to access Header

[petar-m, 2011-02-15 23:46:25]

Hello rsc (cc: golang-dev@googlegroups.com),

I'd like you to review this change to
https://go.googlecode.com/hg/

[mattn, 2011-02-16 01:02:20]

http://codereview.appspot.com/4185053/diff/2001/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/2001/src/pkg/http/response.go#newc...
src/pkg/http/response.go:169: func GetFirstValue(header map[string][]string, key
string) string {
GetFirstHeaderValue?

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go
File src/pkg/websocket/server.go (right):

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:82: origin_, found := req.Header["Origin"]
Why don't you use:
origin, found := req.Header["Origin"][0]

I can't see that there is a point to reference [1] or else.

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:88: key1_, found := req.Header["Sec-Websocket-Key1"]
same as above.

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:92: key2_, found := req.Header["Sec-Websocket-Key2"]
same as above.

[rsc, 2011-02-16 04:27:47]

I started doing this a few weeks ago but got sidetracked.
Thanks for looking at it.  I'll review the CL tomorrow.

Pet peeve without looking at the code, just reacting to the
discussion: "Get" is a Java word.  It is necessary in
Java because you can't have a field named foo and
a method named foo in the same class.  It is not
necessary in Go: because of the export rules the field
is typically named foo and the method Foo.

Any time you think you need a GetFoo method, just
call it Foo instead.  It's shorter, clearer, and easier.

Russ

[petar-m, 2011-02-16 07:07:18]

This is a fair suggestion. Maybe for this or the next CL. Let's see what Russ
thinks.

—Petar
On Tuesday, February 15, 2011 at 9:56 PM, gary b wrote: 
> I recommend adding a type similar to Twister's HeaderMap
(http://garyburd.github.com/twister/pkg/web.html#HeaderMap) and declaring
headers as type HeaderMap instead of map[string][]string. Twister's HeaderMap
has methods for working with multi-value headers. Here are some examples where
HeaderMap will improve the code in this changeset:
> 
> Several instances of this pattern in request.go:
> 
> ct_, ok := r.Header["Content-Type"]
> var ct string
> if ok { 
> ct = ct_[0]
> } 
> 
> can be replaced with:
> 
> ct, ok := r.Header.Get("Content-Type")
> 
> The functionality in the new functions:
> 
> func (r *Response) GetHeaderFirst(key string) string
> GetFirstValue(header map[string][]string, key string) string
> 
> is available in a HeaderMap method:
> 
> func (m HeaderMap) GetDef(key string, def string) string
> 
> HeaderMap is a better place for this functionality because the functionality
is useful outside the context of HTTP responses.
> 
> 
> 
> 

[rsc, 2011-02-16 23:04:43]

Thanks for cleaning this up.


I agree with the suggestion about having a Header type, like:
---
// A Header represents a MIME-style header mapping
// keys to sets of values.
type Header map[string][]string

// Add adds the key, value pair to the header.
// It appends to any existing values associated with key.
func (h Header) Add(key, value string)

// Set sets the header entries associated with key to
// the single element value.  It replaces any existing
// values associated with key.
func (h Header) Set(key, value string)

// Get gets the first value associated with the given key.
// If there are no values associated with the key, Get returns "".
// Get is a convenience method.  For more complex queries,
// access the map directly.
func (h Header) Get(key string) string

// Del deletes the values associated with key.
func (h Header) Del(key string)
---

It probably belongs in textproto but I would be happy to hear
counterarguments.


This will mean that the Trailer field has type Header but so be it.


When you write a header literal you don't need to echo the []string
in the nested composite literal anymore:

h := Header{
	"Cookie": {"xxx", "yyy"},
}


Please avoid the _ suffix on variable names.

[petar-m, 2011-02-17 00:41:30]

I've already implemented the cookie mechanism, factoring your concerns.

It is coming in the CL right after this one.

--P 
—Petar
On Wednesday, February 16, 2011 at 7:18 PM, mattn wrote: 
> 
> > When you write a header literal you don't need to echo the []string
> > in the nested composite literal anymore:
> > h := Header{
> > "Cookie": {"xxx", "yyy"},
> > }
> 
> Sounds good, And I hope Cookie include expire/domain/path/secure/HttpOnly like
below.
> 
> https://github.com/mattn/go-session-manager/blob/master/session.go#L13
> 
> Currently, go's header is joined with comma for cookie.
> Then we should parse Set-Cookie like below.
> 
> https://github.com/mattn/go-session-manager/blob/master/session.go#L155
> 
> If possible, I want that it parse Cookies in http package.
> 
> Thanks.
> - Yasuhiro Matsumoto
> 
> 

[petar-m, 2011-02-17 16:57:01]

textproto.MIMEHeader created

http: adjusted for change

http://codereview.appspot.com/4185053/diff/2001/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/2001/src/pkg/http/response.go#newc...
src/pkg/http/response.go:169: func GetFirstValue(header map[string][]string, key
string) string {
This moved as a method of textproto.Header.

On 2011/02/16 01:02:21, mattn wrote:
> GetFirstHeaderValue?

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go
File src/pkg/websocket/server.go (right):

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:82: origin_, found := req.Header["Origin"]
Because it may generate a panic if the map does not
have an entry for this key.

This is now replaced with a use of a convenience method
of textproto.Header

On 2011/02/16 01:02:21, mattn wrote:
> Why don't you use:
> origin, found := req.Header["Origin"][0]
> 
> I can't see that there is a point to reference [1] or else.

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:88: key1_, found := req.Header["Sec-Websocket-Key1"]
On 2011/02/16 01:02:21, mattn wrote:
> same as above.

Done.

http://codereview.appspot.com/4185053/diff/2001/src/pkg/websocket/server.go#n...
src/pkg/websocket/server.go:92: key2_, found := req.Header["Sec-Websocket-Key2"]
On 2011/02/16 01:02:21, mattn wrote:
> same as above.

Done.

[petar-m, 2011-02-17 16:57:16]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[petar-m, 2011-02-17 17:11:50]

Also, I think that

textproto.CanonicalHeaderKey should be renamed to
textproto.CanonicalMIMEKey

On 2011/02/17 16:57:16, petar-m wrote:
> Hello rsc, mattn (cc: mailto:golang-dev@googlegroups.com),
> 
> Please take another look.

[rsc, 2011-02-17 19:17:07]

The new MIMEHeader code in textproto looks good.

Bunch of comments about usage in http below.  It says
textproto.MIMEHeader in all my comments but the more
I see it the more I think http should define its own header type.
Clients of http shouldn't need to know that http is using
textproto in its implementation, and the amount of code
duplicated is small.

I suggest copying (not moving) the MIMEHeader type
and methods into http and then renaming it to just Header.
Then in my comments below s/textproto.MIMEHeader/Header/g

The general cleanups (use Get and Del wherever possible)
apply to more code than I identified; I stopped at response.go.

I agree about renaming textproto.CanonicalHeaderKey to
textproto.CanonicalMIMEHeaderKey.  i think the one in
http (there should be one in http for the same reason 
as http.Header) should stay as CanonicalHeaderKey.

Thanks for doing this.  I know it's a fair amount of busy work.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go
File src/pkg/http/client.go (left):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#oldcode34
src/pkg/http/client.go:34: // matchNoProxy returns true if requests to addr
should not use a proxy,
I think you ran hg sync and then wrote out your file over
the sync'ed version.  Please restore the proxy code.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go
File src/pkg/http/client.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcode54
src/pkg/http/client.go:54: req.Header = make(map[string][]string)
req.Header = make(textproto.Header)

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcode56
src/pkg/http/client.go:56: req.Header["Authorization"] = []string{"Basic " +
string(encoded)}
might as well use the methods

req.Header.Set("Authorization", "Basic " + string(encoded))

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcod...
src/pkg/http/client.go:171: "Content-Type": []string{bodyType},
req.Header = textproto.MIMEHeader{
    "Content-Type": {bodyType},
}

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcod...
src/pkg/http/client.go:195: req.Header = map[string][]string{
same: use textproto.MIMEHeader, drop []string inside

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go
File src/pkg/http/fs.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go#newcode107
src/pkg/http/fs.go:107: ifmodsince_, ok := r.Header["If-Modified-Since"]
Revert changes here and change the original to just use r.Header.Get.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go#newcode217
src/pkg/http/fs.go:217: func parseRange(s_ []string, size int64) ([]httpRange,
os.Error) {
Revert these.
Change caller to use Header.Get.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go
File src/pkg/http/fs_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:119: req.Header = make(map[string][]string)
make(textproto.MIMEHeader)

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:133: req.Header["Range"] = []string{"bytes=" + rt.r}
req.Header.Set

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:148: cr_, ok := r.Header["Content-Range"]
r.Header.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/readrequest_test.go
File src/pkg/http/readrequest_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/readrequest_test...
src/pkg/http/readrequest_test.go:53: Header: map[string][]string{
usual comment

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/readrequest_test...
src/pkg/http/readrequest_test.go:96: Header:        map[string][]string{},
textproto.MIMEHeader{}

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go
File src/pkg/http/request.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:136: Trailer map[string][]string
Trailer textproto.MIMEHeader

i think?

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:149: v_, ok := r.Header["Content-Type"]
use r.Header.Get.

the v_ for a slice is not an established Go convention either.
but all these variables can go away anyway.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:422: host_, ok := req.Header["Host"]
req.Header.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:427: req.Header["Host"] = nil, false
.Del

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:432: referer_, ok := req.Header["Referer"]
.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:436: req.Header["Referer"] = nil, false
.Del

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:438: useragent_, ok := req.Header["User-Agent"]
.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:442: req.Header["User-Agent"] = nil, false
.Del

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:528: ct_, ok := r.Header["Content-Type"]
.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:560: if vs, ok := r.Form[key]; ok {
the old code was better.  it handled the case where the Form (incorrectly) had
an empty slice stored in the map.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:567: expectation_, ok := r.Header["Expect"]
s/_//

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:575: value_, exists := r.Header["Connection"]
.Get
then test value == ""

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go
File src/pkg/http/request_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go#...
src/pkg/http/request_test.go:78: req.Header = map[string][]string{
usual comments

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go#...
src/pkg/http/request_test.go:100: {contentType: stringMap{"Content-Type":
[]string{"text/plain"}}},
s/[]string//
next few lines too

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go#newc...
src/pkg/http/response.go:133: func fixPragmaCacheControl(header
map[string][]string) {
header Header

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go#newc...
src/pkg/http/response.go:141: // GetFirstHeaderValue returns the first element
of header[key],
delete; let callers use r.Header.Get

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header.go
File src/pkg/net/textproto/header.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:11: // Add adds the key, value pair to the
header.
can delete this method.  no one uses this form as far as i can tell

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:26: func (h MIMEHeader) AddNewLine(key, value
string) {
s/AddNewLine/Add/

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:29: oldValues, ok := h[key]
h[key] = append(h[key], value)

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:52: v_, ok := h[CanonicalHeaderKey(key)]
v := h[CanonicalHeaderKey(key)]
if len(v) == 0 {
    return ""
}
return v[0]

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/reader.go
File src/pkg/net/textproto/reader.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/reader....
src/pkg/net/textproto/reader.go:422: // TODO(petar): Should this method perhaps
split values at commas
No, I think it should just do what it says.
The part about comma-separated things is HTTP-specific.
MIME headers as used in mail don't have this rule.

[petar-m, 2011-02-18 01:18:01]

All done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go
File src/pkg/http/client.go (left):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#oldcode34
src/pkg/http/client.go:34: // matchNoProxy returns true if requests to addr
should not use a proxy,
On 2011/02/17 19:17:08, rsc wrote:
> I think you ran hg sync and then wrote out your file over
> the sync'ed version.  Please restore the proxy code.

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go
File src/pkg/http/client.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcode54
src/pkg/http/client.go:54: req.Header = make(map[string][]string)
On 2011/02/17 19:17:08, rsc wrote:
> req.Header = make(textproto.Header)
> 

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcode56
src/pkg/http/client.go:56: req.Header["Authorization"] = []string{"Basic " +
string(encoded)}
On 2011/02/17 19:17:08, rsc wrote:
> might as well use the methods
> 
> req.Header.Set("Authorization", "Basic " + string(encoded))

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcod...
src/pkg/http/client.go:171: "Content-Type": []string{bodyType},
On 2011/02/17 19:17:08, rsc wrote:
> req.Header = textproto.MIMEHeader{
>     "Content-Type": {bodyType},
> }
> 

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/client.go#newcod...
src/pkg/http/client.go:195: req.Header = map[string][]string{
On 2011/02/17 19:17:08, rsc wrote:
> same: use textproto.MIMEHeader, drop []string inside

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go
File src/pkg/http/fs.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go#newcode107
src/pkg/http/fs.go:107: ifmodsince_, ok := r.Header["If-Modified-Since"]
On 2011/02/17 19:17:08, rsc wrote:
> Revert changes here and change the original to just use r.Header.Get.

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs.go#newcode217
src/pkg/http/fs.go:217: func parseRange(s_ []string, size int64) ([]httpRange,
os.Error) {
On 2011/02/17 19:17:08, rsc wrote:
> Revert these.
> Change caller to use Header.Get.

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go
File src/pkg/http/fs_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:119: req.Header = make(map[string][]string)
On 2011/02/17 19:17:08, rsc wrote:
> make(textproto.MIMEHeader)

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:133: req.Header["Range"] = []string{"bytes=" + rt.r}
On 2011/02/17 19:17:08, rsc wrote:
> req.Header.Set

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/fs_test.go#newco...
src/pkg/http/fs_test.go:148: cr_, ok := r.Header["Content-Range"]
On 2011/02/17 19:17:08, rsc wrote:
> r.Header.Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/readrequest_test.go
File src/pkg/http/readrequest_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/readrequest_test...
src/pkg/http/readrequest_test.go:53: Header: map[string][]string{
On 2011/02/17 19:17:08, rsc wrote:
> usual comment

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go
File src/pkg/http/request.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:136: Trailer map[string][]string
On 2011/02/17 19:17:08, rsc wrote:
> Trailer textproto.MIMEHeader
> 
> i think?

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:149: v_, ok := r.Header["Content-Type"]
On 2011/02/17 19:17:08, rsc wrote:
> use r.Header.Get.
> 
> the v_ for a slice is not an established Go convention either.
> but all these variables can go away anyway.

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:422: host_, ok := req.Header["Host"]
On 2011/02/17 19:17:08, rsc wrote:
> req.Header.Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:427: req.Header["Host"] = nil, false
On 2011/02/17 19:17:08, rsc wrote:
> .Del

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:432: referer_, ok := req.Header["Referer"]
On 2011/02/17 19:17:08, rsc wrote:
> .Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:436: req.Header["Referer"] = nil, false
On 2011/02/17 19:17:08, rsc wrote:
> .Del

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:438: useragent_, ok := req.Header["User-Agent"]
On 2011/02/17 19:17:08, rsc wrote:
> .Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:442: req.Header["User-Agent"] = nil, false
On 2011/02/17 19:17:08, rsc wrote:
> .Del

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:528: ct_, ok := r.Header["Content-Type"]
On 2011/02/17 19:17:08, rsc wrote:
> .Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:560: if vs, ok := r.Form[key]; ok {
On 2011/02/17 19:17:08, rsc wrote:
> the old code was better.  it handled the case where the Form (incorrectly) had
> an empty slice stored in the map.

I don't remember changing this. Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:567: expectation_, ok := r.Header["Expect"]
On 2011/02/17 19:17:08, rsc wrote:
> s/_//

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request.go#newco...
src/pkg/http/request.go:575: value_, exists := r.Header["Connection"]
On 2011/02/17 19:17:08, rsc wrote:
> .Get
> then test value == ""

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go
File src/pkg/http/request_test.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go#...
src/pkg/http/request_test.go:78: req.Header = map[string][]string{
On 2011/02/17 19:17:08, rsc wrote:
> usual comments

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/request_test.go#...
src/pkg/http/request_test.go:100: {contentType: stringMap{"Content-Type":
[]string{"text/plain"}}},
On 2011/02/17 19:17:08, rsc wrote:
> s/[]string//
> next few lines too

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go#newc...
src/pkg/http/response.go:133: func fixPragmaCacheControl(header
map[string][]string) {
On 2011/02/17 19:17:08, rsc wrote:
> header Header

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/http/response.go#newc...
src/pkg/http/response.go:141: // GetFirstHeaderValue returns the first element
of header[key],
On 2011/02/17 19:17:08, rsc wrote:
> delete; let callers use r.Header.Get

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header.go
File src/pkg/net/textproto/header.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:11: // Add adds the key, value pair to the
header.
On 2011/02/17 19:17:08, rsc wrote:
> can delete this method.  no one uses this form as far as i can tell

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:26: func (h MIMEHeader) AddNewLine(key, value
string) {
On 2011/02/17 19:17:08, rsc wrote:
> s/AddNewLine/Add/

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:29: oldValues, ok := h[key]
On 2011/02/17 19:17:08, rsc wrote:
> h[key] = append(h[key], value)

Isn't this going to panic if key is not present in h? Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/header....
src/pkg/net/textproto/header.go:52: v_, ok := h[CanonicalHeaderKey(key)]
On 2011/02/17 19:17:08, rsc wrote:
> v := h[CanonicalHeaderKey(key)]
> if len(v) == 0 {
>     return ""
> }
> return v[0]

Done.

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/reader.go
File src/pkg/net/textproto/reader.go (right):

http://codereview.appspot.com/4185053/diff/3017/src/pkg/net/textproto/reader....
src/pkg/net/textproto/reader.go:422: // TODO(petar): Should this method perhaps
split values at commas
On 2011/02/17 19:17:08, rsc wrote:
> No, I think it should just do what it says.
> The part about comma-separated things is HTTP-specific.
> MIME headers as used in mail don't have this rule.

Done.

[petar-m, 2011-02-18 01:22:53]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[petar-m, 2011-02-18 01:30:20]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[mattn, 2011-02-18 01:39:44]

http://codereview.appspot.com/4185053/diff/13001/src/pkg/http/client.go
File src/pkg/http/client.go (left):

http://codereview.appspot.com/4185053/diff/13001/src/pkg/http/client.go#oldco...
src/pkg/http/client.go:93: var proxyURL *URL
It seems that proxy support was dropped.

[petar-m, 2011-02-18 12:33:17]

Done.

http://codereview.appspot.com/4185053/diff/13001/src/pkg/http/client.go
File src/pkg/http/client.go (left):

http://codereview.appspot.com/4185053/diff/13001/src/pkg/http/client.go#oldco...
src/pkg/http/client.go:93: var proxyURL *URL
On 2011/02/18 01:39:44, mattn wrote:
> It seems that proxy support was dropped.

Done.

[petar-m, 2011-02-18 12:33:33]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[mattn, 2011-02-18 13:26:56]

That's not the point, I got test fail. This is relative problem issue 4177042.

  http://codereview.appspot.com/4177042/

Currently, fs_test.go provide http://localhost/ServeFile, but server return
header "Localtion: \ServeFile".

rsc, Shall we fix 4177042 in this time?
Of course, I should check another pass about manipurating URL and local path.

Thanks.
- Yasuhiro Matsumoto

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go
File src/pkg/http/serve_test.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go#n...
src/pkg/http/serve_test.go:202: t.Errorf("Get(%q) = %q, want %q", vt.url, s_[0],
vt.expected)
if ok is false, s_ will be nil. Then it will be panic.

[mattn, 2011-02-18 13:31:06]

I got PASS with 4177042 just now.

On 2011/02/18 13:26:56, mattn wrote:
> That's not the point, I got test fail. This is relative problem issue 4177042.
> 
>   http://codereview.appspot.com/4177042/
> 
> Currently, fs_test.go provide http://localhost/ServeFile, but server return
> header "Localtion: \ServeFile".
> 
> rsc, Shall we fix 4177042 in this time?
> Of course, I should check another pass about manipurating URL and local path.
> 
> Thanks.
> - Yasuhiro Matsumoto
> 
> http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go
> File src/pkg/http/serve_test.go (right):
> 
>
http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go#n...
> src/pkg/http/serve_test.go:202: t.Errorf("Get(%q) = %q, want %q", vt.url,
s_[0],
> vt.expected)
> if ok is false, s_ will be nil. Then it will be panic.

[rsc, 2011-02-18 13:52:59]

> rsc, Shall we fix 4177042 in this time?
> Of course, I should check another pass about manipurating URL and local
> path.

The code in this CL is fine.
I will take care of the path problems next week.

Russ

[rsc, 2011-02-22 21:53:44]

This looks great.  I like how this turned out.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go
File src/pkg/http/client.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:136: conn, err = tls.Dial("tcp", "", addr, nil)
still missing proxy code
(look at diff)

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:142: h = h[0:strings.LastIndex(h, ":")]
unnecessary change

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:144: if err := conn.(*tls.Conn).VerifyHostname(h); err !=
nil {
unnecessary change

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go
File src/pkg/http/header.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newcode9
src/pkg/http/header.go:9: // A Header represents the key-multivalue pairs in
s/multi//
unwrap - plenty of room.

// A Header represents the key-value pairs in an HTTP header.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newco...
src/pkg/http/header.go:11: type Header textproto.MIMEHeader
s/textproto.MIMEHeader/map[string][]string

so that people reading docs don't have to chase it down

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newco...
src/pkg/http/header.go:40: // MIME header key s.  The canonicalization converts
the first
s/MIME //

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response.go#new...
src/pkg/http/response.go:210: kva := make([]string, 0, len(kvm))
I'm not sure that this adaptation of the original code
is correct.  It gives up the order in the value []string,
which we might as well preserve.

keys := make([]string, 0, len(kvm))
for k := range kvm {
    if !exclude[k] {
        keys = append(keys, k)
    }
}
sort.SortStrings(keys)
for _, k := range keys {
    for _, v := range kvm[k] {
        if _, err := fmt.Fprintf(w, "%s: %s\r\n", k, v); err != nil {
            return err
        }
    }
}
return nil

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response_test.go
File src/pkg/http/response_test.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response_test.g...
src/pkg/http/response_test.go:131: Header:           map[string][]string{},
Look for map[string][]string{} throughout
and replace with Header{}.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go
File src/pkg/http/serve_test.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go#n...
src/pkg/http/serve_test.go:200: s_, ok := r.Header["Result"]
s := r.Header.Get("Result")

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go
File src/pkg/http/transfer.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:24: Trailer          map[string][]string
why not Trailer Header?

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:162: Header        map[string][]string
why not Header Header?

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:172: Trailer          map[string][]string
why not Trailer Header?

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:292: func fixTransferEncoding(header
map[string][]string) ([]string, os.Error) {
header Header

http://codereview.appspot.com/4185053/diff/16001/src/pkg/net/textproto/reader.go
File src/pkg/net/textproto/reader.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/net/textproto/reader...
src/pkg/net/textproto/reader.go:418: //		"My-Key": []string{"Value 1", "Value
2"},
drop []string 
next line too

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go
File src/pkg/websocket/server.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:83: if len(origin) == 0 {
probably origin == ""
using len lets the reader (incorrectly) think it is a slice.
next two times too

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:142: if len(protocol) > 0 {
same: protocol != ""

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:181: if len(origin) == 0 {
origin == ""

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:210: if len(protocol) > 0 {
protocol != ""

[petar-m, 2011-02-22 23:18:44]

K.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go
File src/pkg/http/client.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:136: conn, err = tls.Dial("tcp", "", addr, nil)
On 2011/02/22 21:53:44, rsc wrote:
> still missing proxy code
> (look at diff)
> 

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:142: h = h[0:strings.LastIndex(h, ":")]
On 2011/02/22 21:53:44, rsc wrote:
> unnecessary change

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/client.go#newco...
src/pkg/http/client.go:144: if err := conn.(*tls.Conn).VerifyHostname(h); err !=
nil {
On 2011/02/22 21:53:44, rsc wrote:
> unnecessary change

All these are some weird diff confusion. I thought I had the proxy code in, and
never made those two little changes. Oh well.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go
File src/pkg/http/header.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newcode9
src/pkg/http/header.go:9: // A Header represents the key-multivalue pairs in
On 2011/02/22 21:53:44, rsc wrote:
> s/multi//
> unwrap - plenty of room.
> 
> // A Header represents the key-value pairs in an HTTP header.

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newco...
src/pkg/http/header.go:11: type Header textproto.MIMEHeader
On 2011/02/22 21:53:44, rsc wrote:
> s/textproto.MIMEHeader/map[string][]string
> 
> so that people reading docs don't have to chase it down

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/header.go#newco...
src/pkg/http/header.go:40: // MIME header key s.  The canonicalization converts
the first
On 2011/02/22 21:53:44, rsc wrote:
> s/MIME //

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response.go
File src/pkg/http/response.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response.go#new...
src/pkg/http/response.go:210: kva := make([]string, 0, len(kvm))
On 2011/02/22 21:53:44, rsc wrote:
> I'm not sure that this adaptation of the original code
> is correct.  It gives up the order in the value []string,
> which we might as well preserve.
> 
> keys := make([]string, 0, len(kvm))
> for k := range kvm {
>     if !exclude[k] {
>         keys = append(keys, k)
>     }
> }
> sort.SortStrings(keys)
> for _, k := range keys {
>     for _, v := range kvm[k] {
>         if _, err := fmt.Fprintf(w, "%s: %s\r\n", k, v); err != nil {
>             return err
>         }
>     }
> }
> return nil

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response_test.go
File src/pkg/http/response_test.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/response_test.g...
src/pkg/http/response_test.go:131: Header:           map[string][]string{},
On 2011/02/22 21:53:44, rsc wrote:
> Look for map[string][]string{} throughout
> and replace with Header{}.

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go
File src/pkg/http/serve_test.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go#n...
src/pkg/http/serve_test.go:200: s_, ok := r.Header["Result"]
On 2011/02/22 21:53:44, rsc wrote:
> s := r.Header.Get("Result")
> 

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/serve_test.go#n...
src/pkg/http/serve_test.go:202: t.Errorf("Get(%q) = %q, want %q", vt.url, s_[0],
vt.expected)
On 2011/02/18 13:26:57, mattn wrote:
> if ok is false, s_ will be nil. Then it will be panic.

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go
File src/pkg/http/transfer.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:24: Trailer          map[string][]string
On 2011/02/22 21:53:44, rsc wrote:
> why not Trailer Header?

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:162: Header        map[string][]string
On 2011/02/22 21:53:44, rsc wrote:
> why not Header Header?

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:172: Trailer          map[string][]string
On 2011/02/22 21:53:44, rsc wrote:
> why not Trailer Header?

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/http/transfer.go#new...
src/pkg/http/transfer.go:292: func fixTransferEncoding(header
map[string][]string) ([]string, os.Error) {
On 2011/02/22 21:53:44, rsc wrote:
> header Header

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/net/textproto/reader.go
File src/pkg/net/textproto/reader.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/net/textproto/reader...
src/pkg/net/textproto/reader.go:418: //		"My-Key": []string{"Value 1", "Value
2"},
On 2011/02/22 21:53:44, rsc wrote:
> drop []string 
> next line too

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go
File src/pkg/websocket/server.go (right):

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:83: if len(origin) == 0 {
OK. I did it as a speed optimization.
Is that sort of thing meant to be optimized by the compiler?

On 2011/02/22 21:53:44, rsc wrote:
> probably origin == ""
> using len lets the reader (incorrectly) think it is a slice.
> next two times too

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:142: if len(protocol) > 0 {
On 2011/02/22 21:53:44, rsc wrote:
> same: protocol != ""

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:181: if len(origin) == 0 {
On 2011/02/22 21:53:44, rsc wrote:
> origin == ""

Done.

http://codereview.appspot.com/4185053/diff/16001/src/pkg/websocket/server.go#...
src/pkg/websocket/server.go:210: if len(protocol) > 0 {
On 2011/02/22 21:53:44, rsc wrote:
> protocol != ""

Done.

[petar-m, 2011-02-22 23:18:45]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc, 2011-02-22 23:22:09]

LGTM

sorry about the proxy diff trouble.

[rsc, 2011-02-22 23:26:29]

Please change CL description to

	http: introduce Header type, implement with net/textproto
	textproto: introduce Header type
	websocket: use new interface to access Header

I would have done this during submit but I am unable
to submit this.

It appears that some changes to
net/textproto/reader_test are missing?

Russ

[petar-m, 2011-02-22 23:44:32]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[petar-m, 2011-02-22 23:47:19]

—Petar
On Tuesday, February 22, 2011 at 6:26 PM, Russ Cox wrote: 
> Please change CL description to
> 
>  http: introduce Header type, implement with net/textproto
>  textproto: introduce Header type
>  websocket: use new interface to access Header
> 
> I would have done this during submit but I am unable
> to submit this.
> 
> It appears that some changes to
> net/textproto/reader_test are missing?
> 
I'm not sure what you mean?

Russ
> 

[petar-m, 2011-02-22 23:49:15]

Oh, got it. Added reader_test.go

On 22 February 2011 18:44,  <petarm@gmail.com> wrote:
> Hello rsc, mattn (cc: golang-dev@googlegroups.com),
>
> Please take another look.
>
>
> http://codereview.appspot.com/4185053/
>

[petar-m, 2011-02-22 23:49:17]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc, 2011-02-22 23:53:56]

I think reader_test.go is missing from the CL.

Russ
On Feb 22, 2011 6:47 PM, "Petar Maymounkov" <petarm@gmail.com> wrote:
>
>
> —Petar
> On Tuesday, February 22, 2011 at 6:26 PM, Russ Cox wrote:
>> Please change CL description to
>>
>> http: introduce Header type, implement with net/textproto
>> textproto: introduce Header type
>> websocket: use new interface to access Header
>>
>> I would have done this during submit but I am unable
>> to submit this.
>>
>> It appears that some changes to
>> net/textproto/reader_test are missing?
>>
> I'm not sure what you mean?
>
> Russ
>>

[rsc, 2011-02-23 04:04:49]

I'm sorry to keep bouncing this back to you, but I ran
all.bash and it doesn't pass with this CL patched in.


make[2]: Leaving directory `/home/rsc/g/go/src/pkg/http'
--- FAIL: http.TestRequestWrite (0.0 seconds)
	Test 0, expecting:
	GET http://www.techcrunch.com/ HTTP/1.1
	Host: www.techcrunch.com
	User-Agent: Fake
	Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
	Accept-Encoding: gzip,deflate
	Accept-Language: en-us,en;q=0.5
	Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
	Keep-Alive: 300
	Proxy-Connection: keep-alive
	
	
	Got:
	GET http://www.techcrunch.com/ HTTP/1.1
	Host: www.techcrunch.com
	User-Agent: Fake
	Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
	Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
	Accept-Encoding: gzip,deflate
	Accept-Language: en-us,en;q=0.5
	Keep-Alive: 300
	Proxy-Connection: keep-alive
	
	
FAIL

Could you try running all.bash on your machine?

Thanks.
Russ

[petar-m, 2011-02-23 04:30:52]

Yep ... I'll take care, no worries.

--P


On 22 February 2011 23:04, Russ Cox <rsc@golang.org> wrote:
> I'm sorry to keep bouncing this back to you, but I ran
> all.bash and it doesn't pass with this CL patched in.
>
>
> make[2]: Leaving directory `/home/rsc/g/go/src/pkg/http'
> --- FAIL: http.TestRequestWrite (0.0 seconds)
>        Test 0, expecting:
>        GET http://www.techcrunch.com/ HTTP/1.1
>        Host: www.techcrunch.com
>        User-Agent: Fake
>        Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
>        Accept-Encoding: gzip,deflate
>        Accept-Language: en-us,en;q=0.5
>        Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
>        Keep-Alive: 300
>        Proxy-Connection: keep-alive
>
>
>        Got:
>        GET http://www.techcrunch.com/ HTTP/1.1
>        Host: www.techcrunch.com
>        User-Agent: Fake
>        Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
>        Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
>        Accept-Encoding: gzip,deflate
>        Accept-Language: en-us,en;q=0.5
>        Keep-Alive: 300
>        Proxy-Connection: keep-alive
>
>
> FAIL
>
> Could you try running all.bash on your machine?
>
> Thanks.
> Russ
>

[petar-m, 2011-02-23 05:14:42]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc, 2011-02-23 05:19:53]

gotest works in http and websocket
but it fails in net/textproto

$ gotest
rm -f _test/net/textproto.a _gotest_.6
6g -o _gotest_.6 header.go pipeline.go reader.go textproto.go
writer.go  reader_test.go writer_test.go
rm -f _test/net/textproto.a
gopack grc _test/net/textproto.a _gotest_.6
--- FAIL: textproto.TestReadMIMEHeader (0.0 seconds)
	ReadMIMEHeader: map[Long-Key:[Even Longer Value] My-Key:[Value 1
Value 2]], <nil>; want map[Long-Key:[Even Longer Value] My-Key:[Value
1 Value 2]]
FAIL
$

[petar-m, 2011-02-23 05:25:09]

Hello rsc, mattn (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc, 2011-02-23 05:38:47]

LGTM

Nice catch in the net/textproto fix.
I had no idea what was wrong.

Russ

[rsc, 2011-02-23 05:39:31]

*** Submitted as http://code.google.com/p/go/source/detail?r=8876943e2918 ***

http: introduce Header type, implement with net/textproto
textproto: introduce Header type
websocket: use new interface to access Header

R=rsc, mattn
CC=golang-dev
http://codereview.appspot.com/4185053

Committer: Russ Cox <rsc@golang.org>