RE: further to store & message flow discussion

From: Robert Collins <robert.collins@dont-contact.us>
Date: Tue, 20 Feb 2001 08:42:45 +1100

> -----Original Message-----
> From: Henrik Nordstrom [mailto:hno@hem.passagen.se]
> Sent: Tuesday, 20 February 2001 8:48 AM
> To: Robert Collins
> Cc: squid-dev@squid-cache.org
> Subject: Re: further to store & message flow discussion
>
>
> Robert Collins wrote:
>
> > From what I can see they write to a store entry via
> storeAppend. This buffers arbitrary sizes in memory. The
> creation and reads are
> > async.
>
> Yes, sort of. But it does not need to be that way. If you look at the
> big sources (the object listings and such) you will se that it is done
> incrementally today.
>
> The change of data flow will have effects on the cachemgr functions as
> well. Those will know (by being called) when the requestor can accept
> more data.
>
> The only buffering REQUIRED by the cachemgr functions is the amount
> which cannot easily be broken up in incremental steps.
>
> /Henrik
>

Yes. Thus we need _some_ buffering capability in the intermediary.

Rob
Received on Mon Feb 19 2001 - 14:47:56 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:33 MST