On 3/9/06, Michele Simionato <[EMAIL PROTECTED]> wrote: > I did some simple things in cherrypy, but I am worried. It seems to > me that it has been very fragile in the past and possibly it is > fragile even now. However, it has progressed a lost in the last > two years, where Quixote has remained nearly static. OTOH, there > was nothing to improve in Quixote, whereas there was a lot to > improve in CherryPy.
CherryPy 2.2 RC 1 was released today, so 2.2 final is just around the corner. A documentation update is promised this weekend. I don't know whether it's rock solid now but it's probably less fragile than it was, given the number of programmers who are using it and peering at the source. The main problem with CherryPy vis-a-vis Quixote is that it's a bigger and more ambitious product, so it has more ways to break. > The thing I do not like about CherryPy is the usage of threads. I've been ambivalent about its threads, but my problem with SQLite's thread bug recently did scare me away somewhat. (A three-table join locked up pysqlite in some C code, and quitting the main server did not quit the threads. Upgrading SQLite for a threading bugfix caused a different and fatal error. I'll see if upgrading pysqlite helps. That was the thing that finally pushed me to try Durus.) > I like Twisted deferreds philosophy, but I think a lot of people > will not use it because of its the excessive complexity for simple things.. > > Today I wrote a small module (<50 lines) to make Twisted look like CherryPy > and to get the CherryPy 'Hello World' to run. It is cheating, since it > does not provide the full CherryPy functionality, but it gives the > essential. Unfortunately, in order to write it, you have to spend > ddomr time in order to understand how Twisted works :-( Twisted used to have a WSGI interface that worked with CherryPy, but it broke during the Twisted 2 development and was removed. That was a year ago; maybe it's been replaced by now. The generator enhancements in Python 2.5 will supposedly make Twisted programming easier, since you'll be able to use a single generator with normal top-down programming rather than callbacks. But it requires injecting data (the deferred result) into the generator between each yield, and you can't do that now without abusing shared mutable variables. -- Mike Orr <[EMAIL PROTECTED]> ([EMAIL PROTECTED] address is semi-reliable) _______________________________________________ Quixote-users mailing list [email protected] http://mail.mems-exchange.org/mailman/listinfo/quixote-users
