Re: Yet another feature request for the web-server

[Eli Barzilay <eli@barzilay.org>]

On Aug 31, Francisco Solsona wrote:
> So, the feature I'd like to see is a parameter, say:
>   webserver-output-port
> with the obvious value, so that I could send extra headers to the
> client, before returning the Xexpr the webserver is waiting for.

I think that a much better approach is to have a parameter that
accumulates output headers that you can use to add output lines to.
Accessing the raw output stream might make many people shoot their
feet.  This is the approach I used in my webserver which began its
life in the distant past as a hack of the PLT server (the tricky part
is getting a parameter that has a single value across a single
connection, without having shared parameters (which disappeared ages
ago)).


On Aug 31, MJ Ray wrote:
> I'm not sure.  I think a lot of the point of the webserver is to
> avoid the kind of session management reinvention that cookies are
> usually used for, with the send/suspend way of working.  [...]

It's not only session management -- what about user preferences?  What
about playing with other headers like having a page that will not
stick to your cache, changing mime types etc etc etc?


BTW, I never mentioned my server until now (partly because I don't
think that implementing my own thing is the right thing), but I have
many many things that I needed that made my thing pretty different.  I
guess it's enough to watch the result at
  http://sardine.cs.columbia.edu:8080/pics/
and think about what's needed to implement it to add much more to the
current thing.  (Also, a chance to see Shriram's door as it looked
last year...)

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!