httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From (Robert S. Thau)
Subject Re: patch 90g
Date Mon, 12 Feb 1996 01:11:36 GMT
Hmmm.... just on general design principles, we might want to be a
little careful about *how* scripts requirest a particular buffer size
(if we do give them that capability).  Specifically, it might be good
to adopt some sort of a naming convention for this, and any other new
CGI-specific headers we create, which distinguishes them from whatever
headers are to be added to the MIME headers of the server's response.  
I believe one of the suggestions for what to name the thing was
CGI-buffer-size: --- and in fact, prefixing all new CGI-specific
headers with "CGI-" would probably suit the purpose just fine.

There are two reasons why I think it might be a good idea to follow
such a convention:

1) It would minimize the probability of name collisions between
   headers having some effect on the CGI mechanism, and headers
   to be added to the MIME headers of the server's response.
   (There is currently no syntactic distinction betweeen these
   two cases).

2) It would allow an aging server to detect the case of a script which
   is *trying* to use some new CGI feature which it does not support,
   and treat the thing appropriately (e.g., by *not* forwarding it
   to the client).

Of course, this particular suggestion relies on the HTTP standardization
effort to *not* adopt header names which conflict with whatever convention
we finally choose, but for the moment, we're relying on that anyway; it
just gives them a clear patch of namespace to avoid.  (Is this sensible?

(If not, one alternative might be to adopt a syntax for CGI-specific
headers which is not legal RFC822 header syntax at all, perhaps by using
header names with blanks in them --- I think that's illegal --- viz:

   Content-type: text/html
   Set-cookie: items=grommet,1;frammis,3
   CGI output buffer size: 50000
   CGI log comment: {{Session XYZPDQ} {Action add} {Item frammis} {Qty 3}}

Of course deliberately adopting a syntax for new CGI headers which is
not legal by RFC822 might well be considered an evil in and of itself,
in some quarters... I'm just noting that it might be the lesser of two
evils, if we can't eliminate the possibility of conflict any other way).


View raw message