Add HTTP server mojo interfaces.

Add HTTP server mojo interfaces.

The reasons for adding such interfaces:
- There'll be code in src/mandoline that creates an http server which enables
communication between html_viewer/mandoline ui and external applications. One
use case is to support telemetry.

- We would like to use the C++ http server and websocket impl in net/, and don't
want to link net/ into anything other than the network service.

BUG=478249
TEST=None

Committed: https://crrev.com/50826e07c1cad99bd1dc5984a6ea6ccdd6ddb600
Cr-Commit-Position: refs/heads/master@{#329338}

[yzshen1, 2015-05-08 20:52:50]

yzshen@chromium.org changed reviewers:
+ jam@chromium.org

[yzshen1, 2015-05-08 20:52:51]

Hi,

John: Would you please review? Thanks!
This CL copies http_{request,response}.mojom from the Mojo repo; the rest is
new.

James: FYI since we have discussed about this topic.
It seems simpler and easier to maintain to let the network service provide
interfaces for HTTP server. With this we could consider removing the standalone
http_server in the Mojo repo later.

[jam, 2015-05-11 14:58:40]

On 2015/05/08 20:52:51, yzshen1 wrote:
> Hi,
> 
> John: Would you please review? Thanks!
> This CL copies http_{request,response}.mojom from the Mojo repo; the rest is
> new.

Can you provide more background to make sure I (and others reading this)
understand the motivation? Is this needed because there'll be code in
src/mandoline that looks for a command line flag and creates an http server
which passes commands to html_viewer and mandoline ui? and this needs to use
network service because we'll use the net C++ http server, and don't want to
link net into anything other than the network service?

[yzshen1, 2015-05-11 15:38:08]

On 2015/05/11 14:58:40, jam wrote:
> On 2015/05/08 20:52:51, yzshen1 wrote:
> > Hi,
> > 
> > John: Would you please review? Thanks!
> > This CL copies http_{request,response}.mojom from the Mojo repo; the rest is
> > new.
> 
> Can you provide more background to make sure I (and others reading this)
> understand the motivation? Is this needed because there'll be code in
> src/mandoline that looks for a command line flag and creates an http server
> which passes commands to html_viewer and mandoline ui? and this needs to use
> network service because we'll use the net C++ http server, and don't want to
> link net into anything other than the network service?

Thanks for the suggestion. (I think you have already summarized the motivation
well.)

- There'll be code in src/mandoline that creates an http server which enables
communication between html_viewer/mandoline ui and external applications. One
use case is to support telemetry.

- We would like to use the C++ http server and websocket impl in net/, and don't
want to link net/ into anything other than the network service.

[jam, 2015-05-11 20:54:46]

thanks for the explanation, I wasn't sure.

https://codereview.chromium.org/1136773002/diff/1/mojo/services/network/publi...
File mojo/services/network/public/interfaces/http_request.mojom (right):

https://codereview.chromium.org/1136773002/diff/1/mojo/services/network/publi...
mojo/services/network/public/interfaces/http_request.mojom:7: struct HttpRequest
{
I'm curious is there a reason why we can't reuse URLRequest here?
https://code.google.com/p/chromium/codesearch#chromium/src/mojo/services/netw....

if we can't, headers should be an array of a struct per URLRequest since there
are bugs if a map is used (I just learnt this recently). same for response

[yzshen1, 2015-05-11 22:45:26]

https://codereview.chromium.org/1136773002/diff/1/mojo/services/network/publi...
File mojo/services/network/public/interfaces/http_request.mojom (right):

https://codereview.chromium.org/1136773002/diff/1/mojo/services/network/publi...
mojo/services/network/public/interfaces/http_request.mojom:7: struct HttpRequest
{
On 2015/05/11 20:54:46, jam wrote:
> I'm curious is there a reason why we can't reuse URLRequest here?
>
https://code.google.com/p/chromium/codesearch#chromium/src/mojo/services/netw....
> 
> if we can't, headers should be an array of a struct per URLRequest since there
> are bugs if a map is used (I just learnt this recently). same for response

Good point. Thanks!

[jam, 2015-05-12 01:13:07]

lgtm

[yzshen1, 2015-05-12 02:23:04]

The CQ bit was checked by yzshen@chromium.org

[commit-bot: I haz the power, 2015-05-12 02:24:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1136773002/20001

[commit-bot: I haz the power, 2015-05-12 03:14:49]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-05-12 03:15:34]

Patchset 2 (id:??) landed as
https://crrev.com/50826e07c1cad99bd1dc5984a6ea6ccdd6ddb600
Cr-Commit-Position: refs/heads/master@{#329338}