Robert Collins wrote:
> > There must be something I do not get, but what in your small change
> > made Range processing so much more difficult than it was, given the
> > "old" approach?
>
> At a guess, as I fought with this in te quite heavily... the old range
> code occurred all on the client side, so the stream was the resulting
> ranges, whereas now the range process checking should occur both in the
> cache (can part be satisfied from cache) and the server side (validate
> any cached results, and retrieve the non-available data). Then the
> resulting data is pushed back to through to the client, but as a series
> of offset -length pairs, rather than a stream of data.
I am talking about what we have changed in HEAD now, not where we ideally
want to end up..
The only change I can see in HEAD on the store API so far is that
storeClientCopy now only one one offset argument with slightly different
semantics from before. I don't really see anything in this change which would
prevent the old range processing code from working.
Regards
Henrik
Received on Thu Apr 04 2002 - 06:02:41 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:57 MST