net: move openssl and keygen from net/base to net/cert

net: move openssl_client_key_store from net/base to net/ssl

This is a follow-up to https://codereview.chromium.org/13006020/

BUG=70818
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=191853

[Paweł Hajdan Jr., 2013-03-29 22:43:31]

If the moves look OK, I'd do the usual fix-ups (sorting gyp, windows #include
order, etc).

[Ryan Sleevi, 2013-03-29 22:51:17]

<keygen> is not related to certificates

[Ryan Sleevi, 2013-03-29 22:53:59]

On 2013/03/29 22:51:17, Ryan Sleevi wrote:
> <keygen> is not related to certificates

Which is to say, although <keygen> can be used to generate a key that later has
a certificate delivered to it, it doesn't rely on any of the certificate
handling.

It's not quite a low-level algorithm, which is why it doesn't fit into crypto/,
but it probably doesn't belong in cert either.

The same applies to private_key_store. The client_key_store *is* related
specifically to certificates.

[Paweł Hajdan Jr., 2013-03-29 23:00:55]

On 2013/03/29 22:53:59, Ryan Sleevi wrote:
> On 2013/03/29 22:51:17, Ryan Sleevi wrote:
> > <keygen> is not related to certificates
> 
> Which is to say, although <keygen> can be used to generate a key that later
has
> a certificate delivered to it, it doesn't rely on any of the certificate
> handling.
> 
> It's not quite a low-level algorithm, which is why it doesn't fit into
crypto/,
> but it probably doesn't belong in cert either.

Now keygen is in net/base and depends on moved openssl pieces.

Please suggest an alternative place for keygen, or OK the dependency of net/base
on things outside (ugly).

> The same applies to private_key_store. The client_key_store *is* related
> specifically to certificates.

Could you translate that to specific file moves?

[Ryan Sleevi, 2013-03-29 23:03:05]

On 2013/03/29 23:00:55, Paweł Hajdan Jr. wrote:
> On 2013/03/29 22:53:59, Ryan Sleevi wrote:
> > On 2013/03/29 22:51:17, Ryan Sleevi wrote:
> > > <keygen> is not related to certificates
> > 
> > Which is to say, although <keygen> can be used to generate a key that later
> has
> > a certificate delivered to it, it doesn't rely on any of the certificate
> > handling.
> > 
> > It's not quite a low-level algorithm, which is why it doesn't fit into
> crypto/,
> > but it probably doesn't belong in cert either.
> 
> Now keygen is in net/base and depends on moved openssl pieces.

Not if you don't move the private_key_store

> 
> Please suggest an alternative place for keygen, or OK the dependency of
net/base
> on things outside (ugly).
> 
> > The same applies to private_key_store. The client_key_store *is* related
> > specifically to certificates.
> 
> Could you translate that to specific file moves?

private_key_store stays in net/base
client_key_store moves

[Paweł Hajdan Jr., 2013-03-29 23:37:34]

Done. Fix-ups still to be done, just wanted to make sure this is good.

[Ryan Sleevi, 2013-03-30 00:10:57]

Looking closely, I'm torn between whether this belongs in net/cert or net/ssl

The reason I say this is because net/ssl is where you moved the
client_cert_store files, and this class is only used by the
net/socket/ssl_client_socket_openssl_*, and offers the logical counterpart to
that (but reflecting the different needs of Android)

Let's go with net/ssl for now - I'd much rather keep symmetry of underlying
functionality - and iff this needs to become generic, move it to net/cert.

Don't forget to update the description, sort, etc.

[Paweł Hajdan Jr., 2013-04-01 18:48:13]

PTAL

I've sorted net.gyp and updated commit message. The Rietveld issue title is
preserved so that thread view in gmail can still work. :)

[Ryan Sleevi, 2013-04-01 19:41:15]

lgtm

[Paweł Hajdan Jr., 2013-04-02 17:42:16]

Committed patchset #3 manually as r191853.