Pass the final URL an app was loaded from to the app in Initialize().

Pass the final URL an app was loaded from to the app in Initialize().

R=davemoore@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/81b48458928f45dbcaa3fca7972ab971c089611a

[Aaron Boodman, 2015-02-05 19:54:42]

aa@chromium.org changed reviewers:
+ davemoore@chromium.org, sky@chromium.org

[Aaron Boodman, 2015-02-05 19:54:42]

dave: pls review
sky: fyi

[sky, 2015-02-05 23:41:22]

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
File mojo/public/interfaces/application/application.mojom (right):

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
mojo/public/interfaces/application/application.mojom:18: Initialize(Shell shell,
string url, array<string>? args);
Is there a reason we don't want to pass the url as a chain? What if there was
redirects or other things, might the urls along the path be interesting?

[Aaron Boodman, 2015-02-05 23:45:27]

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
File mojo/public/interfaces/application/application.mojom (right):

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
mojo/public/interfaces/application/application.mojom:18: Initialize(Shell shell,
string url, array<string>? args);
On 2015/02/05 23:41:22, sky wrote:
> Is there a reason we don't want to pass the url as a chain? What if there was
> redirects or other things, might the urls along the path be interesting?

I guess since we don't know the shape of the sandbox yet (origin-shaped?
url-shaped? something else?) we don't know what information makes sense to share
with the app.

And since the final URL seems like all that is needed right now, I thought it
made sense to be conservative.

[sky, 2015-02-05 23:49:51]

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
File mojo/public/interfaces/application/application.mojom (right):

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
mojo/public/interfaces/application/application.mojom:18: Initialize(Shell shell,
string url, array<string>? args);
On 2015/02/05 23:45:27, Aaron Boodman wrote:
> On 2015/02/05 23:41:22, sky wrote:
> > Is there a reason we don't want to pass the url as a chain? What if there
was
> > redirects or other things, might the urls along the path be interesting?
> 
> I guess since we don't know the shape of the sandbox yet (origin-shaped?
> url-shaped? something else?) we don't know what information makes sense to
share
> with the app.
> 
> And since the final URL seems like all that is needed right now, I thought it
> made sense to be conservative.

Doesn't the array<string> provide you with more flexibility for the future? It
can be one for now, but it's easy to make it contain more in the future.

As to the specific case, IMO we should share all and restrict that in the future
if necessary.

[Aaron Boodman, 2015-02-06 00:11:54]

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
File mojo/public/interfaces/application/application.mojom (right):

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
mojo/public/interfaces/application/application.mojom:18: Initialize(Shell shell,
string url, array<string>? args);
On 2015/02/05 23:49:51, sky wrote:
> On 2015/02/05 23:45:27, Aaron Boodman wrote:
> > On 2015/02/05 23:41:22, sky wrote:
> > > Is there a reason we don't want to pass the url as a chain? What if there
> was
> > > redirects or other things, might the urls along the path be interesting?
> > 
> > I guess since we don't know the shape of the sandbox yet (origin-shaped?
> > url-shaped? something else?) we don't know what information makes sense to
> share
> > with the app.
> > 
> > And since the final URL seems like all that is needed right now, I thought
it
> > made sense to be conservative.
> 
> Doesn't the array<string> provide you with more flexibility for the future? It
> can be one for now, but it's easy to make it contain more in the future.
> 
> As to the specific case, IMO we should share all and restrict that in the
future
> if necessary.

AFAIK, after release, it would not be possible to restrict because interfaces
have to be backward compatible.

True, we are a long way away from that. But another reason that I prefer to just
pass the single URL is so that if there is a reason to add more of the URLs in
the future, we will find out about it and have a discussion, rather than just
silently growing dependencies on parts of an API that we weren't sure we should
have.

[Aaron Boodman, 2015-02-06 00:27:56]

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
File mojo/public/interfaces/application/application.mojom (right):

https://codereview.chromium.org/905583002/diff/1/mojo/public/interfaces/appli...
mojo/public/interfaces/application/application.mojom:18: Initialize(Shell shell,
string url, array<string>? args);
On 2015/02/06 00:11:54, Aaron Boodman wrote:
> On 2015/02/05 23:49:51, sky wrote:
> > On 2015/02/05 23:45:27, Aaron Boodman wrote:
> > > On 2015/02/05 23:41:22, sky wrote:
> > > > Is there a reason we don't want to pass the url as a chain? What if
there
> > was
> > > > redirects or other things, might the urls along the path be interesting?
> > > 
> > > I guess since we don't know the shape of the sandbox yet (origin-shaped?
> > > url-shaped? something else?) we don't know what information makes sense to
> > share
> > > with the app.
> > > 
> > > And since the final URL seems like all that is needed right now, I thought
> it
> > > made sense to be conservative.
> > 
> > Doesn't the array<string> provide you with more flexibility for the future?
It
> > can be one for now, but it's easy to make it contain more in the future.
> > 
> > As to the specific case, IMO we should share all and restrict that in the
> future
> > if necessary.
> 
> AFAIK, after release, it would not be possible to restrict because interfaces
> have to be backward compatible.
> 
> True, we are a long way away from that. But another reason that I prefer to
just
> pass the single URL is so that if there is a reason to add more of the URLs in
> the future, we will find out about it and have a discussion, rather than just
> silently growing dependencies on parts of an API that we weren't sure we
should
> have.

With a moment more thought (and double checking with Yuzhu) I realized that
changing the type either way would not be possible without breaking backward
compat.

But it doesn't matter because we would use the ordinal system to rename the old
method to InitializeDeprecated and add a new method with the new params named
Initialize.

So moving the interface either direction -- to more permissive or less
permissive will work.

It's true that just having an array now would be less work in the future, but by
that logic we should just make all our types array<string>, or just JavaScript
:).

[Aaron Boodman, 2015-02-06 23:31:01]

aa@chromium.org changed reviewers:
+ hans@chromium.org

[Aaron Boodman, 2015-02-06 23:31:02]

+hans for js changes

[hans, 2015-02-06 23:32:36]

hans@chromium.org changed reviewers:
+ hansmuller@chromium.org

[hans, 2015-02-06 23:32:37]

> +hans for js changes

s/hans/hansmuller/

[hansmuller1, 2015-02-06 23:41:56]

On 2015/02/06 23:32:37, hans wrote:
> > +hans for js changes
> 
> s/hans/hansmuller/

The JS changes look fine to HansMuller.

[DaveMoore, 2015-02-07 00:06:00]

lgtm

[Aaron Boodman, 2015-02-10 20:24:10]

New patchsets have been uploaded after l-g-t-m from davemoore@chromium.org

[Aaron Boodman, 2015-02-10 21:54:27]

Committed patchset #10 (id:180001) manually as
81b48458928f45dbcaa3fca7972ab971c089611a (presubmit successful).