Measure inappropriate_fallback alerts from Google.

Measure inappropriate_fallback alerts from Google.

Some fraction of successful connections use the fallback, but only due to
a spurious network failure. To estimate this fraction, compare handshakes
to Google servers which succeed against those that fail with an
inappropriate_fallback alert. Google servers are known to implement
FALLBACK_SCSV, so a spurious network failure while connecting would
trigger the fallback, successfully connect, but fail with this alert.

BUG=459690
Committed: https://crrev.com/701ca98bb1301a5e90f878aca6ca036a509fcd00
Cr-Commit-Position: refs/heads/master@{#330425}

[davidben, 2015-05-15 01:01:33]

davidben@chromium.org changed reviewers:
+ rsleevi@chromium.org

[davidben, 2015-05-15 01:01:34]

I put this on top of the other CL, so try bots are unhappy with the patch.

[Ryan Sleevi, 2015-05-15 22:24:43]

lgtm

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
File net/http/http_network_transaction.cc (right):

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
net/http/http_network_transaction.cc:1455: // instead meet with an
inappropriate_fallback alert. Comparing occurences
spelling: occurrences

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
net/http/http_network_transaction.cc:1456: // of OK and
ERR_SSL_INAPPROPRIATE_FALLBACK estimates spurious fallbacks.
I find this sentence a bit of an incomplete thought - comparing them doesn't
intrinsically estimate, although it allows you to. But it may help to elaborate
precisely how it allows you to (briefly). Or just omitting this and making sure
it's captured in the CL description.

[davidben, 2015-05-15 23:27:56]

davidben@chromium.org changed reviewers:
+ asvitkine@chromium.org

[davidben, 2015-05-15 23:27:57]

+asvitkine for histograms.xml

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
File net/http/http_network_transaction.cc (right):

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
net/http/http_network_transaction.cc:1455: // instead meet with an
inappropriate_fallback alert. Comparing occurences
On 2015/05/15 22:24:43, Ryan Sleevi wrote:
> spelling: occurrences

Done.

https://codereview.chromium.org/1147453003/diff/1/net/http/http_network_trans...
net/http/http_network_transaction.cc:1456: // of OK and
ERR_SSL_INAPPROPRIATE_FALLBACK estimates spurious fallbacks.
On 2015/05/15 22:24:43, Ryan Sleevi wrote:
> I find this sentence a bit of an incomplete thought - comparing them doesn't
> intrinsically estimate, although it allows you to. But it may help to
elaborate
> precisely how it allows you to (briefly). Or just omitting this and making
sure
> it's captured in the CL description.

I think it's worthwhile to have it in source, since people don't always git
blame. How's this new version?

[Ryan Sleevi, 2015-05-15 23:52:17]

lgtm

[davidben, 2015-05-18 17:42:04]

davidben@chromium.org changed reviewers:
+ isherman@chromium.org

[davidben, 2015-05-18 17:42:05]

+isherman for histograms.xml (looks like it's a holiday in Canada today?)

[Ilya Sherman, 2015-05-18 18:55:34]

LGTM

https://codereview.chromium.org/1147453003/diff/40001/tools/metrics/histogram...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/1147453003/diff/40001/tools/metrics/histogram...
tools/metrics/histograms/histograms.xml:48281: +  <int value="1"
label="inappropriate_fallback alert"/>
nit: Why not "Inappropriate fallback alert"?  (Which I'd personally further
tweak to "Inappropriate fallback used", but that's mostly a personal
preference.)

[davidben, 2015-05-18 19:09:09]

https://codereview.chromium.org/1147453003/diff/40001/tools/metrics/histogram...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/1147453003/diff/40001/tools/metrics/histogram...
tools/metrics/histograms/histograms.xml:48281: +  <int value="1"
label="inappropriate_fallback alert"/>
On 2015/05/18 18:55:34, Ilya Sherman wrote:
> nit: Why not "Inappropriate fallback alert"?  (Which I'd personally further
> tweak to "Inappropriate fallback used", but that's mostly a personal
> preference.)

Alerts are a type of error in the TLS protocol and inappropriate_fallback is the
name of the alert. The server is sending us the inappropriate_fallback alert.

[davidben, 2015-05-18 19:46:25]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2015-05-18 19:53:59]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1147453003/40001

[commit-bot: I haz the power, 2015-05-18 21:21:55]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-05-18 22:43:40]

Patchset 3 (id:??) landed as
https://crrev.com/701ca98bb1301a5e90f878aca6ca036a509fcd00
Cr-Commit-Position: refs/heads/master@{#330425}