Allow the number of undecryptable QUIC packets buffered before the

Allow the number of undecryptable QUIC packets buffered before the
CHLO/SHLO to be configured in QuicConfig.

Merge internal change: 76707156

On chromium side: set max undecryptable packets to the initial
congestion window size in QuicStreamFactory.

R=rch@chromium.org

[Ryan Hamilton, 2014-10-10 22:51:10]

rch@chromium.org changed reviewers:
+ ianswett@google.com

[Ryan Hamilton, 2014-10-10 22:51:11]

IIRC, I think we also want to make a chrome-side change to set the max
undecryptable packets to something like the initial congestion window size in
the QuicConfig. (I think this logic is in the QuicStreamFactory). Your merge
LGTM, but we should also add change the value chrome uses.

[ramant (doing other things), 2014-10-10 23:19:52]

On 2014/10/10 22:51:11, Ryan Hamilton wrote:
> IIRC, I think we also want to make a chrome-side change to set the max
> undecryptable packets to something like the initial congestion window size in
> the QuicConfig. (I think this logic is in the QuicStreamFactory). Your merge
> LGTM, but we should also add change the value chrome uses.

Thanks very much Ryan. Updated the QuicStreamFactory's QuicConfig's
max_undecryptable_packets.

[ramant (doing other things), 2014-10-10 23:40:59]

Patchset #2 (id:110001) has been deleted

[ramant (doing other things), 2014-10-10 23:45:16]

On 2014/10/10 23:19:52, ramant wrote:
> On 2014/10/10 22:51:11, Ryan Hamilton wrote:
> > IIRC, I think we also want to make a chrome-side change to set the max
> > undecryptable packets to something like the initial congestion window size
in
> > the QuicConfig. (I think this logic is in the QuicStreamFactory). Your merge
> > LGTM, but we should also add change the value chrome uses.
> 
> Thanks very much Ryan. Updated the QuicStreamFactory's QuicConfig's
> max_undecryptable_packets.

Hi Ryan,
  Set the max_undecryptable_packets_ to either 32 (https) or 20 (http) same as
initial congestion window size. I hope I got it right. PTAL.

thanks
raman

[Ian Swett, 2014-10-11 03:47:44]

On 2014/10/10 23:45:16, ramant wrote:
> On 2014/10/10 23:19:52, ramant wrote:
> > On 2014/10/10 22:51:11, Ryan Hamilton wrote:
> > > IIRC, I think we also want to make a chrome-side change to set the max
> > > undecryptable packets to something like the initial congestion window size
> in
> > > the QuicConfig. (I think this logic is in the QuicStreamFactory). Your
merge
> > > LGTM, but we should also add change the value chrome uses.
> > 
> > Thanks very much Ryan. Updated the QuicStreamFactory's QuicConfig's
> > max_undecryptable_packets.
> 
> Hi Ryan,
>   Set the max_undecryptable_packets_ to either 32 (https) or 20 (http) same as
> initial congestion window size. I hope I got it right. PTAL.
> 
> thanks
> raman

In case we ever want to increase initial CWND for some connections, I would
suggest a larger value, such as 100, unless there is a concern that would cause
problems for Chrome.

[ramant (doing other things), 2014-10-13 17:43:35]

On 2014/10/11 03:47:44, Ian Swett wrote:
> On 2014/10/10 23:45:16, ramant wrote:
> > On 2014/10/10 23:19:52, ramant wrote:
> > > On 2014/10/10 22:51:11, Ryan Hamilton wrote:
> > > > IIRC, I think we also want to make a chrome-side change to set the max
> > > > undecryptable packets to something like the initial congestion window
size
> > in
> > > > the QuicConfig. (I think this logic is in the QuicStreamFactory). Your
> merge
> > > > LGTM, but we should also add change the value chrome uses.
> > > 
> > > Thanks very much Ryan. Updated the QuicStreamFactory's QuicConfig's
> > > max_undecryptable_packets.
> > 
> > Hi Ryan,
> >   Set the max_undecryptable_packets_ to either 32 (https) or 20 (http) same
as
> > initial congestion window size. I hope I got it right. PTAL.
> > 
> > thanks
> > raman
> 
> In case we ever want to increase initial CWND for some connections, I would
> suggest a larger value, such as 100, unless there is a concern that would
cause
> problems for Chrome.

Done.

thanks Ian,
raman