Re: html prefetching

I somehow looked into this issue before. Let me throw my two cents in. ^_~

In my opinion, it is good to do pre-fetching but we probably shouldn't do it
too aggressively. Based on my limited knowledge of HTML, I believe it would be
good if we pre-fetch those objects referenced by SRC="" because the objects
referenced by SRC will be automatically requested without user's input (well,
they can still abort them with STOP button ~.~). So pre-fetching them should
be very beneficial in most cases. However, for those HREF tags, they are only
referenced when the user click on the link. Hence, unless you have a very
intelligent prediction algorithm (or co-operation from the web server, ref
[Using Predictive Prefetching to Improve WWW latency, Venkata Padmanabhan and
Jeff Mogul), it hardly worths it. And if we can really make nice predictions,
we probably should work in the Wall Street, right? ;)

Another issue related to this is that we need to parse the HTML document. To
do this, I presume it is safer and easier to parse it when we have the whole
page. That means we can't keep passing around 4K buffers and discard them when
we send it back to the client. (I only hacked 2.2STABLE5, let me know if I am
wrong here) . We have to keep all the 4K trunks of an HTML document in memory
and parse it before sending the whole thing back to the browser. This also
gives us some extra time to pre-fetch SRC referenced objects. That's how I
think it can be done, is there a better way to do it?

Cheers!
Yee Man
> Andres Kroonmaa wrote:
>
> > On 3 Jun 2000, at 11:16, Alex Rousskov <rousskov@ircache.net> wrote:
> >
> > > Those who believe that the answer a sure YES,
> > > may want to ask themselves a simple question: how come nobody but
> > > CacheFlow has implemented and promoting that feature?
> > Well, all your other arguments are very serious, but not this one.
> > If we always asked ourselves such questions, I'd be still climbing
> > the trees, I guess ;P Someone is always the first, does this by
> > default mean that they must be wrong? Or was that a tricky question?
>
> I'll add another footnote here. All the cacheflow installations I know have t
his
> feature turned off. (No offense to the CacheFlow salesbots who've been buggin
g
> me for the last couple months)
>
> D
Received on Thu Jun 29 2000 - 10:50:59 MDT