Cache flushing/maintenance page. was Re: creating new users in roller 2.0

[Anil Gangolli <an...@busybuddha.org>]

OK.  I have fixes for the bookmark/folder/referrer lack of invalidation 
bugs in my workspace.  I'll check them in if there are no objections, 
and if someone promises to review them.  That's not saying there aren't 
more missing invalidations, and I'm inclined to support retaining the 
maintenance page.  As a side note, it also lets you force a rebuild of 
the search index.

--a.

Anil Gangolli wrote:

> Just confirmed we still have the bookmark change / no cache 
> invalidation bug anyway.
>
> Anil Gangolli wrote:
>
>>
>> Agree that it should be redundant, but I should say I have needed to 
>> use it in the past.  I think the particular bug I was hitting was 
>> that after changing bookmarks or removing referrers (can't remember 
>> which, may have been both), I used to have to clear the cache to make 
>> the change appear.  Haven't done this recently (just haven't changed 
>> those; rarely add a bookmark, and  I no longer display referrers), 
>> and I know the cache code has changed somewhat in recent versions so 
>> it may have been fixed too. I'll try testing for it.
>>
>> It is really just a CYA measure for bugs where we somewhere forget to 
>> invalidate the cache, but we may have users that were working around 
>> something like me (shame on me) rather than reporting it (or in my 
>> case, hunting it down).
>>
>> --a.
>>
>>
>> Dave Johnson wrote:
>>
>>> I'm not sure this worth the effort to add back in. Users should not 
>>> need to flush the cache. If they do, we have a bug. - Dave
>>>
>>>
>>> On Nov 12, 2005, at 10:25 PM, Allen Gilliland wrote:
>>>
>>>> Along these same lines, I had noticed that the user maintenance 
>>>> page went missing in 2.0 as well.  I wasn't sure if this was on 
>>>> purpose, but it is nice to have an easy way for users to clear 
>>>> their cache if they need to.
>>>>
>>>> -- Allen
>>>>
>>>>
>>>>
>

Re: Cache flushing/maintenance page. was Re: creating new users in roller 2.0

[Dave Johnson <da...@rollerweblogger.org>]

I'm checking this now and restoring the maintenance page.

- Dave


On Nov 13, 2005, at 3:06 PM, Anil Gangolli wrote:

>
> I've checked this in as Revision 333105 on the trunk.  Dave, Allen: 
> please review.  Thanks.
>
> Allen Gilliland wrote:
>
>> This would definitely be helpful to me since I am doing the caching 
>> rewrite for Roller 2.1.  I had noticed that we didn't invalidate all 
>> necessary objects and I've been adding them in as I go.  My guess is 
>> that we haven't noticed this because our default cache implementation 
>> uses 60 minute timeouts, which really shouldn't be needed.  I have 
>> included a new non-expiring caching implementation for Roller 2.1 
>> which I think would be a better choice for the default cache, but 
>> obviously we need to make sure we properly invalidate objects before 
>> we use it.
>>
>> -- Allen
>>
>>
>> Anil Gangolli wrote:
>>
>>>
>>> OK.  I have fixes for the bookmark/folder/referrer lack of 
>>> invalidation bugs in my workspace.  I'll check them in if there are 
>>> no objections, and if someone promises to review them.  That's not 
>>> saying there aren't more missing invalidations, and I'm inclined to 
>>> support retaining the maintenance page.  As a side note, it also 
>>> lets you force a rebuild of the search index.
>>>
>>> --a.
>>>
>>> Anil Gangolli wrote:
>>>
>>>> Just confirmed we still have the bookmark change / no cache 
>>>> invalidation bug anyway.
>>>>
>>>> Anil Gangolli wrote:
>>>>
>>>>>
>>>>> Agree that it should be redundant, but I should say I have needed 
>>>>> to use it in the past.  I think the particular bug I was hitting 
>>>>> was that after changing bookmarks or removing referrers (can't 
>>>>> remember which, may have been both), I used to have to clear the 
>>>>> cache to make the change appear.  Haven't done this recently (just 
>>>>> haven't changed those; rarely add a bookmark, and  I no longer 
>>>>> display referrers), and I know the cache code has changed somewhat 
>>>>> in recent versions so it may have been fixed too. I'll try testing 
>>>>> for it.
>>>>>
>>>>> It is really just a CYA measure for bugs where we somewhere forget 
>>>>> to invalidate the cache, but we may have users that were working 
>>>>> around something like me (shame on me) rather than reporting it 
>>>>> (or in my case, hunting it down).
>>>>>
>>>>> --a.
>>>>>
>>>>>
>>>>> Dave Johnson wrote:
>>>>>
>>>>>> I'm not sure this worth the effort to add back in. Users should 
>>>>>> not need to flush the cache. If they do, we have a bug. - Dave
>>>>>>
>>>>>>
>>>>>> On Nov 12, 2005, at 10:25 PM, Allen Gilliland wrote:
>>>>>>
>>>>>>> Along these same lines, I had noticed that the user maintenance 
>>>>>>> page went missing in 2.0 as well.  I wasn't sure if this was on 
>>>>>>> purpose, but it is nice to have an easy way for users to clear 
>>>>>>> their cache if they need to.
>>>>>>>
>>>>>>> -- Allen
>>>>>>>
>>>>>>>
>>>>>>>
>>>>
>>>
>>
>>
>

Re: Cache flushing/maintenance page. was Re: creating new users in roller 2.0

[Anil Gangolli <an...@busybuddha.org>]

I've checked this in as Revision 333105 on the trunk.  Dave, Allen: 
please review.  Thanks.

Allen Gilliland wrote:

> This would definitely be helpful to me since I am doing the caching 
> rewrite for Roller 2.1.  I had noticed that we didn't invalidate all 
> necessary objects and I've been adding them in as I go.  My guess is 
> that we haven't noticed this because our default cache implementation 
> uses 60 minute timeouts, which really shouldn't be needed.  I have 
> included a new non-expiring caching implementation for Roller 2.1 
> which I think would be a better choice for the default cache, but 
> obviously we need to make sure we properly invalidate objects before 
> we use it.
>
> -- Allen
>
>
> Anil Gangolli wrote:
>
>>
>> OK.  I have fixes for the bookmark/folder/referrer lack of 
>> invalidation bugs in my workspace.  I'll check them in if there are 
>> no objections, and if someone promises to review them.  That's not 
>> saying there aren't more missing invalidations, and I'm inclined to 
>> support retaining the maintenance page.  As a side note, it also lets 
>> you force a rebuild of the search index.
>>
>> --a.
>>
>> Anil Gangolli wrote:
>>
>>> Just confirmed we still have the bookmark change / no cache 
>>> invalidation bug anyway.
>>>
>>> Anil Gangolli wrote:
>>>
>>>>
>>>> Agree that it should be redundant, but I should say I have needed 
>>>> to use it in the past.  I think the particular bug I was hitting 
>>>> was that after changing bookmarks or removing referrers (can't 
>>>> remember which, may have been both), I used to have to clear the 
>>>> cache to make the change appear.  Haven't done this recently (just 
>>>> haven't changed those; rarely add a bookmark, and  I no longer 
>>>> display referrers), and I know the cache code has changed somewhat 
>>>> in recent versions so it may have been fixed too. I'll try testing 
>>>> for it.
>>>>
>>>> It is really just a CYA measure for bugs where we somewhere forget 
>>>> to invalidate the cache, but we may have users that were working 
>>>> around something like me (shame on me) rather than reporting it (or 
>>>> in my case, hunting it down).
>>>>
>>>> --a.
>>>>
>>>>
>>>> Dave Johnson wrote:
>>>>
>>>>> I'm not sure this worth the effort to add back in. Users should 
>>>>> not need to flush the cache. If they do, we have a bug. - Dave
>>>>>
>>>>>
>>>>> On Nov 12, 2005, at 10:25 PM, Allen Gilliland wrote:
>>>>>
>>>>>> Along these same lines, I had noticed that the user maintenance 
>>>>>> page went missing in 2.0 as well.  I wasn't sure if this was on 
>>>>>> purpose, but it is nice to have an easy way for users to clear 
>>>>>> their cache if they need to.
>>>>>>
>>>>>> -- Allen
>>>>>>
>>>>>>
>>>>>>
>>>
>>
>
>

Re: Cache flushing/maintenance page. was Re: creating new users in roller 2.0

[Allen Gilliland <Al...@Sun.COM>]

This would definitely be helpful to me since I am doing the caching 
rewrite for Roller 2.1.  I had noticed that we didn't invalidate all 
necessary objects and I've been adding them in as I go.  My guess is 
that we haven't noticed this because our default cache implementation 
uses 60 minute timeouts, which really shouldn't be needed.  I have 
included a new non-expiring caching implementation for Roller 2.1 which 
I think would be a better choice for the default cache, but obviously we 
need to make sure we properly invalidate objects before we use it.

-- Allen


Anil Gangolli wrote:

>
> OK.  I have fixes for the bookmark/folder/referrer lack of 
> invalidation bugs in my workspace.  I'll check them in if there are no 
> objections, and if someone promises to review them.  That's not saying 
> there aren't more missing invalidations, and I'm inclined to support 
> retaining the maintenance page.  As a side note, it also lets you 
> force a rebuild of the search index.
>
> --a.
>
> Anil Gangolli wrote:
>
>> Just confirmed we still have the bookmark change / no cache 
>> invalidation bug anyway.
>>
>> Anil Gangolli wrote:
>>
>>>
>>> Agree that it should be redundant, but I should say I have needed to 
>>> use it in the past.  I think the particular bug I was hitting was 
>>> that after changing bookmarks or removing referrers (can't remember 
>>> which, may have been both), I used to have to clear the cache to 
>>> make the change appear.  Haven't done this recently (just haven't 
>>> changed those; rarely add a bookmark, and  I no longer display 
>>> referrers), and I know the cache code has changed somewhat in recent 
>>> versions so it may have been fixed too. I'll try testing for it.
>>>
>>> It is really just a CYA measure for bugs where we somewhere forget 
>>> to invalidate the cache, but we may have users that were working 
>>> around something like me (shame on me) rather than reporting it (or 
>>> in my case, hunting it down).
>>>
>>> --a.
>>>
>>>
>>> Dave Johnson wrote:
>>>
>>>> I'm not sure this worth the effort to add back in. Users should not 
>>>> need to flush the cache. If they do, we have a bug. - Dave
>>>>
>>>>
>>>> On Nov 12, 2005, at 10:25 PM, Allen Gilliland wrote:
>>>>
>>>>> Along these same lines, I had noticed that the user maintenance 
>>>>> page went missing in 2.0 as well.  I wasn't sure if this was on 
>>>>> purpose, but it is nice to have an easy way for users to clear 
>>>>> their cache if they need to.
>>>>>
>>>>> -- Allen
>>>>>
>>>>>
>>>>>
>>
>