Re: lynx-dev sol.slcc.edu problem (HTTP/1.x?)

[Klaus Weide]

On Tue, 19 Jan 1999, Leonid Pauzner wrote:

> Seems may people now experienced problems
> while downloading resent development lynx version, including myself.

The complaint was originally about a missing feature of the ftp daemon.
Now you are talking about the httpd, a completely different topic.

> Something broken at slcc server, it is cache-unfriendly and broke HTTP/1.0
> but claims itself as HTTP/1.1
> I made a test:
> 
> 1) HEAD responce is not equivalent to head of GET responce
> (no Last-Modified: in the first case).

a. HEAD is far more broken in lots of places at lots of sites, for
   example scripts returning a full message with body.
b. What you have shown doesn't seem to happen if you access a URL
   that gives a full filename (not a generated directory listing or
   directory index file).
c. What you observe doesn't break the protocol.  RFC 2068 9.4: 
   "The metainformation contained
   in the HTTP headers in response to a HEAD request SHOULD be identical
   to the information sent in response to a GET request."
   but not MUST.
d. A proxy relying too much (more than required by the protocol) on
   correct header information from HEAD responses would be stupid.
   If the missing Last-Modified header somehow causes squid to uncache
   that page, it's probably squid's fault.

> It is _possible_ the same happen
> when sending a responce on "conditional GET" which affect the proxies.

That's only a wild guess so far.

> 2) The server does not send Etag: which is a BAD idea
> so HTTP/1.1 proxy may try to load the document over the net again.
> This definitely increase the load of the server.

It may be a bad Idea but it is not a BAD idea.  Etag is a nice extra,
but neither Lynx nor most clients are using it AFAIK.  And I doubt
that squid does much with it.  And squid is not a "HTTP/1.1 proxy".

  ----

So sol.slcc.edu is using a http server which may not be the best
according to some criteria, so what?  Scott McGee is contributing the
resources, and it's his choice what server he wants to run, and you
haven't really shown that the choice of server is responsible for
problems you have observed (which you haven't described), and nobody
seems to be offering on lynx-dev to provide a better service, and
there already is a European mirror, and further discussion about the
state of the http server at sol.slcc.edu (or offers for help) would
probably be more productively directed at Scott, and reports about
flaws in the server software at Netscape (or their successors).


   Klaus