Re: Problem with archives -- mail gets lost

On Mon, Oct 12, 2015 at 10:40 AM, Szymon Lipiński <mabewlun(at)gmail(dot)com>
wrote:

>
>
> On 12 October 2015 at 10:15, Magnus Hagander <magnus(at)hagander(dot)net> wrote:
>
>>
>>
>> On Mon, Oct 12, 2015 at 3:43 AM, Peter Geoghegan <
>> peter(dot)geoghegan86(at)gmail(dot)com> wrote:
>>
>>> This is a dead link, but should link to a post to pgsql-hackers sent a
>>> little while ago:
>>>
>>>
>>> http://www.postgresql.org/message-id/CAM3SWZScpWzQ-7EJC77vwqzZ1GO8GNmURQ1QqDQ3wRn7AbW1Cg@mail.gmail.com
>>>
>>> I can even see my mail listed here, with a dead link to above:
>>>
>>> http://www.postgresql.org/list/pgsql-hackers/since/201510120000/
>>>
>>> The subject of the mail is "INSERT ... ON CONFLICT documentation
>>> clean-up patch". What's up with this?
>>>
>>
>> That link works fine for me.
>>
>> Is there any chance that you actually tried navigating to that link
>> *before* the mail actually showed up on the list in your second page? It's
>> normal with a short delay (usually seconds) after the email gets to
>> majordomo before it's in the archives. At that point it should not be
>> visible on either of those two links, but if you hit the first link it
>> would cache the 404 response for a while. Once the email arrived, the
>> second link would be refreshed but not the first once (since at the time
>> you got to the first link initially, we did not know which thread to
>> expire, as we had not seen the mail yet).
>>
>> Could that be the explanation?
>>
>
>
> Hi,
> I clicked on the first link a minute ago, got 404. Clicked on the second,
> seen the email link, clicked, got email. Went to the browser tab with the
> first link, refreshed, and I had the email. Maybe a kind of caching
> problems?
>
>
Yes, it could be - there are three different cache servers (in different
datacenters) that could have slightly different origin times.

We should probably have a (much) shorter cache time for 404 responses on
actual messages -- but I'd first like to figure out how it got in there in
the first place. The scenario outlined above would make that happen - if
that's not what happened it must be something else, but I'm not sure what.
It seems highly unlikely that a search engine for example would randomly
hit a messageid before we actually had it, since there would be nothing
linking to it...

--
Magnus Hagander
Me: http://www.hagander.net/
Work: http://www.redpill-linpro.com/