On Tue, Oct 17, 2000, Henrik Nordstrom wrote:
> Adrian Chadd wrote:
>
> > Ok, imagine that the disk subsystem/FS could be written in such a way
> > that non-cacheable objects/streaming objects don't ever hit the disk.
> > So rather than having the complexity there for *everything* it will
> > only exist for non-disk objects.
>
> Problem is how you reliably detect these infinite objects without first
> going into VM mode for all objects without a content-length header, or
> defer to make such object public until fully cached.
>
> The main problems arise because Squid allows more than one client to a
> object retreival. If this isn't allowed then we can quite easily fiddle
> around with these areas, to for example make a transition from disk to
> non-disk in the middle of a retreival to solve the "infinite object"
> issue.
Ok. I'll start ripping bits and pieces out of the code on modio and
see what happens.
-- Adrian Chadd The Law of Software Development and <adrian@creative.net.au> Envelopment at MIT: "Every program in development at MIT expands until it can read mail."Received on Fri Oct 20 2000 - 00:14:20 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:51 MST