From: mde@...26.com
Date: 2004-12-25T16:38:10+09:00
Subject: Re: [PATCH] cgi.rb -- more GET/POST stuff

Some more thoughts / clarification on cgi.rb. This might be
a bit long ...

While the question of when it's appropriate to use GET versus
POST is interesting (BTW, I'm not sure I agree with the
idea that you should always expect to be able to bookmark or
e-mail URIs -- especially in a Web application environment),
it's not really relevant to the problem I was describing with
cgi.rb. I guess my poor choice of example (the session ID)
is to blame for that.

The problem is that cgi.rb does not allow you to use any query
string variables at all, if you use the POST method. Take, for
example, a URI which is pretty normal for any Front Controller /
Page Controller setup for a Web app:

/processingpage.rbx?module=account

If I try to POST to this page, the initialize_query() function
does not pick up the variables from the query string, so
cgi['module'] is not there.

There are, of course, a couple of workarounds -- e.g.,
submitting hidden form elements or parsing the data myself
off cgi.rb's query_string environment variable. But other
mainstream Web programming languages like PHP or ASP don't
force me to do this -- query string vars show up properly
whether I choose to use GET or POST to request a URI.

Unless there is some specific reason that POSTs should ignore
what's on the query string, it might be nice for Web programmers
moving to Ruby if it were as easy as the other languages they
have used -- and as intuitive as the rest of Ruby.

Regarding Eustaquio's idea of having two separate hashes
for GET and POST, I think it makes good sense too -- PHP and
ASP have similar setups (PHP's $_GET and $_POST, and ASP's
Request.QueryString and Request.Form), so you can distinguish
between the two when you want to. But they also have a unified
hash (PHP's $_REQUEST and ASP's Request) with everything in it.

However, in practice, the only time I've ever actually needed
to specify one or the other was with legacy code where the
developer had made the confusing decision to use identically
named varibles (with two different values), one on the query
string and one in the form. I'm thinking this is probably not
a great idea, and likely not something you should do all the
time, anyway.

If anybody has insight into how other Web programming languages
handle GET/POST/Cookies, that might be helpful to know. I'm just
using PHP and ASP as examples because it's what I know, and
they are really widespread as Web languages.

Hope I didn't put you folks to sleep.

Thanks.


Matthew