On Mon, 03 Oct 2011, Roman Chyla wrote: > I worked extensively with Twisted and Genshi, both of them are nice > tools and smart, Genshi had some learning curve, but Twisted was just > awkward.
We did not consider Twisted actually, since nobody proposed it. As for Genshi, the XML mode was not liked very much; and while one could use a text mode, there is a problem of speed. In my simplistic tests Genshi was 15-20 times slower than Jinja/Mako. An order of magnitude difference is simply too much. Naturally, templating may not always be the bottleneck, but it is not good to be so wasteful. A fast system provides reassurance and ample breathing space if/when the needs come. Indico guys found similar timings, e.g. for a complex room booking calendar page display, Mako took 1.5 sec while Genshi took 6 to 9 sec. One usable for production, the other much less so. > somehow, I tend to judge frameworks based on the quality of their > documentation and especially API documentation +1 Also, the overall presence of developers on the mailing lists, the blogosphere, and stuff, is rather important. Let me cite just one impressive example, Graham Dumpleton (mod_wsgi). Anyway, these `strategical' aspects of the framework selection process were not wikified on our page, but they are definitely noticed and taken into account. > I missed the elements of joy and discovery. The level of joy while working with various potential candidate frameworks is naturally rather subjective. But it is important indeed. I consider joy to be a self-understood criterion behind all our software endeavours anyway. If it wasn't fun, wouldn't we quit already? :) Best regards -- Tibor Simko
