Caching Issues Followup

Right now we have these six bugs which are suspected to be
caching  
related:

http://tinyurl.com/2lwz7a

Travis put a "band-aid" patch on which we think
should stop all these  
bugs from happening - but since they're so hard to reproduce
we don't  
know for sure.

I'm goign to close out these bugs for now, only re-opening
them up  
again if we see these issues again.

If anyone objects please let me know and I'll re-open them.

Bobby
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

That query was wrong: use this one:

http://tinyurl.com/26uqsu


On Jul 25, 2007, at 10:55 AM, Bobby Rullo wrote:

> Right now we have these six bugs which are suspected to
be caching  
> related:
>
> http://tinyurl.com/2lwz7a
>
> Travis put a "band-aid" patch on which we
think should stop all  
> these bugs from happening - but since they're so hard
to reproduce  
> we don't know for sure.
>
> I'm goign to close out these bugs for now, only
re-opening them up  
> again if we see these issues again.
>
> If anyone objects please let me know and I'll re-open
them.
>
> Bobby
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

Bobby Rullo wrote:
> I'm goign to close out these bugs for now, only
re-opening them up again 
> if we see these issues again.

Ok, no objection.  Might be wise to open a future ticket for
"see if we 
can turn any caching back on", if the current direction
in BZ#9715 
doesn't pan out or stay open.

-- Jared

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On 7/25/07, Bobby Rullo <brosafoundation.org>
wrote:
> Right now we have these six bugs which are suspected to
be caching
> related:
>
> http://tinyurl.com/2lwz7a
>
> Travis put a "band-aid" patch on which we
think should stop all these
> bugs from happening - but since they're so hard to
reproduce we don't
> know for sure.
>
> I'm goign to close out these bugs for now, only
re-opening them up
> again if we see these issues again.

the net result is that we hopefully won't see the bugs
anymore, but we
won't be taking advantage of caching either, right? is that
really the
way we want to release 0.7?
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

I agree it's not ideal, but it seems like right now it's a
choice  
between "slower and correct" and "faster and
wrong".

There are other bugs which are blockers  on everyone's plate
and this  
one has proven to be quite a time-suck for Travis (and
myself to a  
lesser extent)

If we had a developer focus on pretty much just this for a
while I'm  
sure we could solve it, but the issue is...what is "a
while"? Do we  
let this hold up preview?

My opinion is that no, we shouldn't hold up preview for this
bug, but  
we should fix it soon after.

bobby

On Jul 25, 2007, at 11:36 AM, Brian Moseley wrote:

> On 7/25/07, Bobby Rullo <brosafoundation.org>
wrote:
>> Right now we have these six bugs which are
suspected to be caching
>> related:
>>
>> http://tinyurl.com/2lwz7a
>>
>> Travis put a "band-aid" patch on which we
think should stop all these
>> bugs from happening - but since they're so hard to
reproduce we don't
>> know for sure.
>>
>> I'm goign to close out these bugs for now, only
re-opening them up
>> again if we see these issues again.
>
> the net result is that we hopefully won't see the bugs
anymore, but we
> won't be taking advantage of caching either, right? is
that really the
> way we want to release 0.7?
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On 7/25/07, Bobby Rullo <brosafoundation.org>
wrote:
> I agree it's not ideal, but it seems like right now
it's a choice
> between "slower and correct" and "faster
and wrong".

only if the option of fixing the broken caching behavior is
taken off
the table, which i think is premature.

> There are other bugs which are blockers  on everyone's
plate and this
> one has proven to be quite a time-suck for Travis (and
myself to a
> lesser extent)

i understand that this bug is hard to chase down, and for
that reason
other blockers should have priority over it, but once
they're done, we
should try again.

> If we had a developer focus on pretty much just this
for a while I'm
> sure we could solve it, but the issue is...what is
"a while"? Do we
> let this hold up preview?

+0. i don't believe we should release without fixing the
bug, but as i
probably can't effectively contribute to finding the
problem, i can't
take a firmer stance.
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

I sympathize with you here. At this point it's a matter of
developer  
resources.

I'm working on the last couple of bugs I have and then
performance  
stuff (outside perf gains that can be obtained through
caching that  
is  )

Matthew has more than his share.

That leaves Travis, who has done a bunch of work in this
area  
already, and who has very little on his plate now.

So with that in mind it seems like Travis could either take
some mde  
bugs or work on this caching stuff.

Matthew, how badly do you need to get some of these bugs off
your plate?

bobby

ps - travis has seen weird date headers come back from Cosmo
he  
thinks. Can't give more details than this, but it's
something for you  
guys to follow up on with each other.

On Jul 25, 2007, at 2:34 PM, Brian Moseley wrote:

> On 7/25/07, Bobby Rullo <brosafoundation.org>
wrote:
>> I agree it's not ideal, but it seems like right now
it's a choice
>> between "slower and correct" and
"faster and wrong".
>
> only if the option of fixing the broken caching
behavior is taken off
> the table, which i think is premature.
>
>> There are other bugs which are blockers  on
everyone's plate and this
>> one has proven to be quite a time-suck for Travis
(and myself to a
>> lesser extent)
>
> i understand that this bug is hard to chase down, and
for that reason
> other blockers should have priority over it, but once
they're done, we
> should try again.
>
>> If we had a developer focus on pretty much just
this for a while I'm
>> sure we could solve it, but the issue is...what is
"a while"? Do we
>> let this hold up preview?
>
> +0. i don't believe we should release without fixing
the bug, but as i
> probably can't effectively contribute to finding the
problem, i can't
> take a firmer stance.
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

>
>> There are other bugs which are blockers  on
everyone's plate and this
>> one has proven to be quite a time-suck for Travis
(and myself to a
>> lesser extent)
>
> i understand that this bug is hard to chase down, and
for that reason
> other blockers should have priority over it, but once
they're done, we
> should try again.
+1
>
>> If we had a developer focus on pretty much just
this for a while I'm
>> sure we could solve it, but the issue is...what is
"a while"? Do we
>> let this hold up preview?
>
> +0. i don't believe we should release without fixing
the bug, but as i
> probably can't effectively contribute to finding the
problem, i can't
> take a firmer stance.
>

+/-0 (does that exist?) with the caveat that I'd feel
stronger (+1)  
if we can determine that fixing this bug would provide
significant  
end user experience (that is, performance) enhancements.

Right now my best idea for a way forward is documented on
bug 9715.  
The short version is that I've determined we most likely do
not see  
this behavior if we explicitly remember and set
Etag/If-None-Match  
and Last-Modified/If-Modified-Since headers. The one caveat
is that  
we see slightly different XHR behavior that I think would
require at  
least a days worth of refactoring, and possibly more if
Safari/IE  
behavior is not the same as Firefox behavior in this
regard.

The other possible way forward would be to start talking to
the  
Firefox developers. As best as I can figure this is in
reality a  
Firefox bug. However, given our current lack of reliable
reproduction  
steps I can only assume this would take long enough as to be
 
definitely out of scope for 0.7.

-Travis
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

>
> ps - travis has seen weird date headers come back from
Cosmo he  
> thinks. Can't give more details than this, but it's
something for  
> you guys to follow up on with each other.

Actually, after further consideration I'm pretty sure the
headers I  
saw were correct. If I see a similar phenomenon again I'll
raise my  
eyebrows, but I'm pretty sure what we were seeing was a
result of

a) using a calendar that hadn't been modified since the day
before and
b) the strange "old headers show in firebug"
behavior we see with  
this bug

-Travis
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On Jul 25, 2007, at 4:36 PM, Travis Vachon wrote:

> The other possible way forward would be to start
talking to the  
> Firefox developers. As best as I can figure this is in
reality a  
> Firefox bug. However, given our current lack of
reliable  
> reproduction steps I can only assume this would take
long enough as  
> to be definitely out of scope for 0.7.

This may indeed be a legitimate FF bug, but the installed
base of FF  
is not going to upgrade in time to be meaningful to us, so
we can't  
rely on that.

Ted
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On Jul 25, 2007, at 4:35 PM, Bobby Rullo wrote:

> I sympathize with you here. At this point it's a matter
of  
> developer resources.
>
> I'm working on the last couple of bugs I have and then
performance  
> stuff (outside perf gains that can be obtained through
caching that  
> is  )
>
> Matthew has more than his share.
>
> That leaves Travis, who has done a bunch of work in
this area  
> already, and who has very little on his plate now.
>
> So with that in mind it seems like Travis could either
take some  
> mde bugs or work on this caching stuff.
>
> Matthew, how badly do you need to get some of these
bugs off your  
> plate?

I swapped IM's w/ MDE here at OSCON - he has 3 bugs that
Travis could  
take:

https://bugzilla.osafoundation.org/show_bug.cgi?id=10073

https://bugzilla.osafoundation.org/show_bug.cgi?id=10113

https://bugzilla.osafoundation.org/show_bug.cgi?id=8500

Can we reliably reproduce the caching bugs?  I thought that
this was  
somewhat intermittent

Ted

>
> bobby
>
> ps - travis has seen weird date headers come back from
Cosmo he  
> thinks. Can't give more details than this, but it's
something for  
> you guys to follow up on with each other.
>
> On Jul 25, 2007, at 2:34 PM, Brian Moseley wrote:
>
>> On 7/25/07, Bobby Rullo <brosafoundation.org> wrote:
>>> I agree it's not ideal, but it seems like right
now it's a choice
>>> between "slower and correct" and
"faster and wrong".
>>
>> only if the option of fixing the broken caching
behavior is taken off
>> the table, which i think is premature.
>>
>>> There are other bugs which are blockers  on
everyone's plate and  
>>> this
>>> one has proven to be quite a time-suck for
Travis (and myself to a
>>> lesser extent)
>>
>> i understand that this bug is hard to chase down,
and for that reason
>> other blockers should have priority over it, but
once they're  
>> done, we
>> should try again.
>>
>>> If we had a developer focus on pretty much just
this for a while I'm
>>> sure we could solve it, but the issue is...what
is "a while"? Do we
>>> let this hold up preview?
>>
>> +0. i don't believe we should release without
fixing the bug, but  
>> as i
>> probably can't effectively contribute to finding
the problem, i can't
>> take a firmer stance.
>> _______________________________________________
>> cosmo-dev mailing list
>> cosmo-devlists.osafoundation.org
>> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev
>
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

Agreed, sorry I was unclear - the short term purpose of
going to the  
FF devs would be to try to trace down exactly what the bug
is in an  
attempt to find a more elegant workaround.

-Travis

On Jul 25, 2007, at 5:45 PM, Ted Leung wrote:

> On Jul 25, 2007, at 4:36 PM, Travis Vachon wrote:
>
>> The other possible way forward would be to start
talking to the  
>> Firefox developers. As best as I can figure this is
in reality a  
>> Firefox bug. However, given our current lack of
reliable  
>> reproduction steps I can only assume this would
take long enough  
>> as to be definitely out of scope for 0.7.
>
> This may indeed be a legitimate FF bug, but the
installed base of  
> FF is not going to upgrade in time to be meaningful to
us, so we  
> can't rely on that.
>
> Ted
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

>
> Can we reliably reproduce the caching bugs?  I thought
that this  
> was somewhat intermittent
>
> Ted
>

Ever since caching was turned off, we haven't seen those
issues. We  
might have to turn it back on to see how we could reproduce
it. We  
know for sure they weren't happening consistently and
reliably but we  
were all seeing some symptons of it in various operations. 
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev