Re: [boost] [GSoC] Opinions and suggestions for improving CGI library

Subject: Re: [boost] [GSoC] Opinions and suggestions for improving CGI library
From: Phil Endecott (spam_from_boost_dev_at_[hidden])
Date: 2009-03-28 08:50:00

Eugene Wee wrote:
> As suggested, I have uploaded my proposal draft to a web page:
> http://www.comp.nus.edu.sg/~weehke/gsoc/proposal2009.html

Hi Eugene,

Quoting from your web page:

> Session handling allows data to persist across requests. Each session 
> has a unique session identifier, and stores an arbitrary number of 
> name/value pairs. The name will be a string, while the value will 
> be a serializable object. The session identifier could be propagated 
> via a cookie, the query string, or a hidden form field. To avoid 
> session fixation, session identifiers can be regenerated.
>
> At the moment, the CGI library lacks session handling entirely. My 
> aim is to design a session handler interface that allows for pluggable 
> storage formats: a (possibly memory mapped) file or a database table 
> could be used to store the session's name/value pairs. As such, this 
> part of the project will involve designing both the session interface 
> and the session handler interface, and then an implementation will 
> be made with at least one possible storage format.

Web applications have persistent data associated with sessions, but 
they also have persistent data associated with e.g. a username or page, 
etc.  I would think that IFF the framework provides a good way to 
access cookies, form variables etc AND handle the data storage 
(database, Boost.Serialisation, Boost.Interprocess) THEN the user 
should be allowed to plumb these together as they wish: there are many 
possible combinations with no "one size fits all".

> Currently, multipart/form-data parsing is done with regular expressions. 
> However, regular expressions are not adequate to describe the structure, 
> hence this is difficult to read and maintain. I intend to re-implement 
> multipart/form-data parsing by using Boost.Spirit.

multipart/form-data is basically MIME, so providing a library that can 
also be used more generally would be useful.  It would be a shame to 
implement a MIME parser and then to hide it as an implementation detail 
inside another library.

I have previously implemented a basic multipart/form-data parser using 
"traditional" string manipulation (i.e. find, substr etc).  It's not 
clear to me that Boost.Spirit offers a significant benefit for this task.

Anyway, this is all important stuff that we should have in Boost.  Go 
for it.

Phil.