lynx-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: lynx-dev dev.16 patch 3 - SOURCE_CACHE etc.


From: Klaus Weide
Subject: Re: lynx-dev dev.16 patch 3 - SOURCE_CACHE etc.
Date: Thu, 9 Dec 1999 21:02:30 -0600 (CST)

On Thu, 9 Dec 1999, Leonid Pauzner wrote:
> 
> Speaking of SOURCE_CACHE, I think it would be nice to allow different
> number of documents for source cache against HText cache. HText keeps
> around 4 (?) times more space allocated compared to the source size, it

Really?  Are you talking about --enable-color-style?

> would be a reasonable savings to reparse source cache under the cost of
> some extra CPU load and (possible) lost of changed form data. Also,
> source cache can be utilized when 'd'ownloading html files if they match
> cache. [different expiration model for anchor->FileCache and
> anchor->source_cache* but at least we could duplicate data without extra
> net load].

I have said in the past that I think the whole source cache thing uses
a fundamentally wrong approach.  I still think so.  Since it more or
less works for what it does, and I haven't provided anything better,
there wasn't much point in belaboring it.

Retrieving from "source cache" should go through getfile() like other
requests, instead of doing its own thing and duplicating stuff that is
done in the normal path.  (I think the history of changes shows that
those shortcuts taken have resulted in lots of problems - things that
happen in the normal path and had to be discovered and duplicated for
"reparsing".)

In the current "shortcut" way of things, a source cache copy can only
reasonably used if there is still a HText object around.  That's just
how it is - and source_cache isn't a "real" cache (in anything like
an HTTP sense).  To make it one would require adding some sensible
cache-control features.  I don't see how that is possible with the
current approach, without adding a lot more "duplication" or potential
for duplication (between the HTreparse_document -> HTParseFoo shortcut
path and the top-of-mainloop -> getfile -> HTLoad* -> protocol path).

A cahce without any cache-control features - one that caches everything
that comes in and never expires etc. - just doesn't make sense, it's
a bloating monster.

What about the good old advice - if you want a real cache, use an
external one.

   Klaus


reply via email to

[Prev in Thread] Current Thread [Next in Thread]