RE: [Axiom-developer] Re: Axiom HyperDoc Replacement

[Bill Page]

On April 21, 2007 3:38 PM Martin Rubey wrote:
> 
> "Bill Page" <address@hidden> writes:
> 
> > Or we could just add a couple of extra lisp commands to
> > open and send the contents of the file ourselves. 
> 
> Well, that's what I have done now -- following the code in 
> the sandboxsockets page.  But it's nowhere as nice as the
> simple redirect.

Nice. Seems ok to me. Since this can be compiled, I expect
it to be pretty fast - for a single user it is probably
faster than the overhead of having more processes running.

> 
> > It's not a big deal. The main thing is that you found a way 
> > to call a Spad routine [...]
> 
> > Do you have an explanation why you want/need to use the 
> > *docfun* name format?
> 
> > Oh, I didn't know that this was a problem.  Hm, you did it 
> > yourself with MySort(L,<$Integer)$Lisp, so I do not really
> > follow?

It did not work they way I originally wrote it. See bottom of
the page:

http://wiki.axiom-developer.org/SandBoxSocketsInSpad

> 
> The only thing that was surprising is that the function needs 
> to be compiled, i.e., defining a function in the interpreter
> does not work, for some reason.

Maybe a limitation of the SPADCALL macro?

> 
> Thus, I modified bar as below. The only reason I made it a 
> global was that I am lazy.  You could just as well pass it
> to server as an argument.

In my code that's what I tried but I did not use it the same
way you did below:

> 
> ;; try it with
> 
> ;; )lisp (load "http-test.lisp")
> ;; BAR(8080, getDocumentation$HyperDoc)$Lisp
> 
> (defvar *docfun*)
> 
> ;; file: http-test.lisp
> (defun bar (p docfun)
>   (setq *docfun* docfun)
>   (let ((s (si::socket p :server #'server)))
>     (tagbody l
>              (when (si::listen s)
>                (let ((w (si::accept s)))
>                  (server w)))
>              (sleep 0.1)
>              (go l))))
> 
> (defun server (s)
>   (let* ((get (read s nil 'eof))
>          (fn (and (eq get 'get) (subseq (string-downcase 
> (read s nil 'eof)) 1))))
> ; subseq removes the leading "/"
>     (format t "Got ~S~%~%" fn)
> ;;; NOTYET tested
>     (when (string= (char fn 1) "?")
>         (setq fn (SPADCALL fn *docfun*)))
> ; *docfun* should return the name of an html file
> ; in case of an error, it will also return the name of an 
> ; html file, which will contain the error message
> ;;; END NOTYET tested
> 
>     (when (string= (subseq fn (- (length fn) 4)) "html")
>         (format s "HTTP/1.1 ~S~%" (if fn 200 403)))
>

; You should probably send an HTTP Content-type: header here
: before the file contents rather than embedding it in the
; HTML file, since that would make the file not readable
; directly by local file access

          (format s "Content-type: text/html~%~%")
 
>     (with-open-file  (q fn) (si::copy-stream q s))
>     (close s)))
> 



> 
> > I recall something about such mangled names being "global" 
> > or in some other way special in scope.
> 
> No, the stars are only a convention.  defvar makes a variable 
> "global".
> 

Ok. Thanks.

> > Is that right? Is that why it works, incontrast to the 
> > lambda function I tried?
> 
> Huh? what lambda function did not work for you?
> 

No, it didn't work. See the bottom of the page. That is the
main reason why I created the page - so I could document this
problem. Maybe it's a bug in GCL? Or maybe it is just I do
not understand how socket function calls server function
internally and that this failure is expected.

> > There is a way to chage the configuration of FireFox:
> > 
> > http://kb.mozillazine.org/Links_to_local_pages_do_not_work
> > 
> > And Konqueror is supposed to just give a warning. Did you
> > see that?
> 
> No, he says access denied.  But in any case, we cannot expect
> a user to fiddle with security settings.

Agreed.

> 
> The "small" webservers, can they be installed without being
> root?
> 

Yes, so long as you assign a non-standard port number >1024 and
don't have an aggressive firewall policy.

Regards,
Bill Page.