QUIC - Always pesist 200 MRU alternate protocols to preferences file.

QUIC - Always pesist 200 MRU alternate protocols to preferences file.

Deleted the experiment to load either 200 or 1000 servers from disk.

+ Obsoleted the following histograms
  ++ Net.AlternateProtocolUsage.200Truncated
  ++ Net.AlternateProtocolUsage.1000Truncated

Reverted the changes related to the above from the following CL:
  https://codereview.chromium.org/298683010/

R=jar@chromium.org, rch@chromium.org

Committed: https://crrev.com/7a6c3e8f4978572660f16b3724c8dbcf0205761c
Cr-Commit-Position: refs/heads/master@{#301070}

[ramant (doing other things), 2014-10-23 21:02:52]

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
File net/http/http_server_properties_manager.cc (right):

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
net/http/http_server_properties_manager.cc:47: const int
kMaxAlternateProtocolHostsToPersist = 1000;
rch and jar: we could reduce it to 200 (which could help with browser startup).
There is a 0.05% difference between persisting 200 vs 1000 MRU servers for the
following two histograms on all platforms. WDYT?

Net.AlternateProtocolUsage.200Truncated
Net.AlternateProtocolUsage.1000Truncated

[jar (doing other things), 2014-10-23 21:38:43]

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
File net/http/http_server_properties_manager.cc (right):

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
net/http/http_server_properties_manager.cc:47: const int
kMaxAlternateProtocolHostsToPersist = 1000;
On 2014/10/23 21:02:52, ramant wrote:
> rch and jar: we could reduce it to 200 (which could help with browser
startup).
> There is a 0.05% difference between persisting 200 vs 1000 MRU servers for the
> following two histograms on all platforms. WDYT?
> 
> Net.AlternateProtocolUsage.200Truncated
> Net.AlternateProtocolUsage.1000Truncated

Sounds like a good proposal.

You could also look at being more lazy in restoring the latter 800, to prevent
impact on startup.  

...but your proposal sure seems small and effective!

[ramant (doing other things), 2014-10-23 22:53:57]

PTAL.

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
File net/http/http_server_properties_manager.cc (right):

https://codereview.chromium.org/675783005/diff/1/net/http/http_server_propert...
net/http/http_server_properties_manager.cc:47: const int
kMaxAlternateProtocolHostsToPersist = 1000;
On 2014/10/23 21:38:43, jar wrote:
> On 2014/10/23 21:02:52, ramant wrote:
> > rch and jar: we could reduce it to 200 (which could help with browser
> startup).
> > There is a 0.05% difference between persisting 200 vs 1000 MRU servers for
the
> > following two histograms on all platforms. WDYT?
> > 
> > Net.AlternateProtocolUsage.200Truncated
> > Net.AlternateProtocolUsage.1000Truncated
> 
> Sounds like a good proposal.
> 
> You could also look at being more lazy in restoring the latter 800, to prevent
> impact on startup.  
> 
> ...but your proposal sure seems small and effective!

Changed the default value to 200.

Done.

[jar (doing other things), 2014-10-23 23:11:51]

https://codereview.chromium.org/675783005/diff/20001/net/http/http_server_pro...
File net/http/http_server_properties_manager.cc (left):

https://codereview.chromium.org/675783005/diff/20001/net/http/http_server_pro...
net/http/http_server_properties_manager.cc:448: if (count >=
alternate_protocols_to_load)
I think this is OK.  Just to confirm.... we will "reload" a potentially larger
list, but we'll only save a smaller (200 element, max) list.  Is that correct?

Should we consider avoiding loading more than some(?) max count?  ... or even
avoid re-loading more than we know we'll next save?

[ramant (doing other things), 2014-10-23 23:27:20]

The only change I have made from Patch Set 1 is to change
kMaxAlternateProtocolHostsToPersist to 200 (backed out the other change in Patch
Set 2). PTAL. Thanks.

https://codereview.chromium.org/675783005/diff/20001/net/http/http_server_pro...
File net/http/http_server_properties_manager.cc (left):

https://codereview.chromium.org/675783005/diff/20001/net/http/http_server_pro...
net/http/http_server_properties_manager.cc:448: if (count >=
alternate_protocols_to_load)
On 2014/10/23 23:11:51, jar wrote:
> I think this is OK.  Just to confirm.... we will "reload" a potentially larger
> list, but we'll only save a smaller (200 element, max) list.  Is that correct?
> 

Made this code similar to other MRU items we are persisting. An admin could edit
preferences file and add more entries (but when we persist the data to
preferences whenever an entry is added or deleted from the cache, it will only
save 200 entries).


> Should we consider avoiding loading more than some(?) max count?  ... or even
> avoid re-loading more than we know we'll next save?

Added back the delete code from previous patch (thought checking the count could
be unnecessary after browser is restarts couple of times).

[jar (doing other things), 2014-10-24 00:39:32]

lgtm

[ramant (doing other things), 2014-10-24 05:44:14]

The CQ bit was checked by rtenneti@chromium.org

[commit-bot: I haz the power, 2014-10-24 05:45:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/675783005/40001

[commit-bot: I haz the power, 2014-10-24 07:26:28]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2014-10-24 07:27:15]

Patchset 3 (id:??) landed as
https://crrev.com/7a6c3e8f4978572660f16b3724c8dbcf0205761c
Cr-Commit-Position: refs/heads/master@{#301070}