Re: upcoming changes to TTLs.

From: Christian Balzer <>
Date: Mon, 28 Oct 1996 19:42:15 +0100 (MET)

Duane Wessels wrote:
> writes:
>>> There will be a performance reduction due to the
>>> latency time on the IMS check with the source.
>>> But it ensure that there will not be a stale object.
>>How about an asynchronous IMS? That is, the IMS is sent to the server
>>while returning the object from the cache (assuming it is still
>>fresh). And this can be done within, say, 20% of the object becoming
>>stale. If the object is already stale, then you have to wait for the
>>IMS. So you don't have to wait for the IMS unless the object is
> suggested the same thing.
And me thought about it for some time, too. ;) ;)

>How would you like to specify this in the config file? global option?
>stuff it in the refresh_pattern line? what are the parameters?
I think the sensible thing would be another pattern thingy, as some would
like 20% others 30% and with some URLs totally different tactics would
be adviseable. I'd like to be able to define a range of staleness IMS
check, like 30% before an object gets stale and 10% afterwards.

In this context I'd like to refine (more like redefine ;-) my idea about
serving stale pages (original site not reachable, etc). If this happens,
just serve it (and use HTTP 1.1 methods to signal a stale copy, if there's
such an option). But if the client does a reload (maybe limited to
pragma no_cache), generate a warning page that the original site can't
be reached for whatever reason but that stale copies can and will be served.
Sensible enuff? ^_^

Mata ne,


  // <CB> aka Christian Balzer, Tannenstr. 23c, D-64342 Seeheim, Germany
\X/ | Voice: +49 6257 83036, Fax/Data: +49 6257 83037
SWB  - The Software Brewery - | Team H, Germany HQ | Anime no Otaku
Received on Mon Oct 28 1996 - 10:50:32 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:33:23 MST