On Wed, Jun 04, 2003 at 01:48:59PM -0700, Ilya Zakharevich wrote:
> As Leonid noticed, we need to set up the purge policy; the current
> (unacceptable) one is to purge before attempting to reload; but
> purging only if the download is successfully completed also looks
> flawed.
>
> What about purging when HEAD is processed?
More remarks from Leonid, answered together:
A) If the document is in HText cache, the expiration date is ignored.
This is not what I observe. Very often documents get reloaded
even with my cache-100-documents settings.
Will try to record the URL when I see this the next time...
B) Honoring timeout is subject to the timeskew between client and server.
Probably; but there is nothing we can do about this, right?
C) no-cache is equivalent to timeout=now.
Do not think so; see time-skew above...
Moreover, documents supplying no-cache are "differently
broken", so it may make sense to treat them differently than
documents with short expiration time.
D) purging on receiving the HEAD is blah blah blah (could not follow
it completely).
Sorry, what I meant was the "header-part" of the document, one
finished by the first "\r\n\r\n" (if I got this correct).
When we receive the header, it should be a good indication
that the server is going to deliver the document - at least
not worse than it did the last time ;-).
This answers the question about 'z', right?
E) We cannot fix major bugs now, when we are close to release.
We are still in development branch. Hurry!!! ;-)
Thanks,
Ilya
; To UNSUBSCRIBE: Send "unsubscribe lynx-dev" to [EMAIL PROTECTED]
