Titto,

I've had no problems with hack. The only things to keep in mind are outside the scope of hack such as:

* Persistence. Clearly you need to optimize your application different for CGI run (load up only what you need right now) versus long-running processes like FastCGI (load data only once).
* URL schemes. A lot of people assume that your web app with be served from the root of the domain. When using my simpleserver testing, that *is* the case. However, I deploy apps in subdirectories of my domain (http://www.snoyman.com/photos/, http://www.snoyman.com/wordify/, etc), so I need to keep this in mind.

My only two quips about hack itself is:

* Versioning scheme. I wish (and have requested) that Hack would follow the Package Versioning Policy so that I could easily check for breaking changes. As is, I simply have to declare the exact version number of Hack I want to work with to guarantee my apps aren't broken in the future.
* More serious issue is that it returns the response as a lazy bytestring. It's not really fair to call this a quip, since I fully supported this approach; nonetheless, using an enumerator for this would probably be more efficient for certain use cases.

I just a few hours ago sent off an e-mail about bringing into fruition the Web Application Interface for Haskell, which I would envision as Hack with these two quips addressed. Theoretically, it would also allow easy collaboration with Hack.

Michael