Hi Norman,
On 8/10/12 4:10 PM, Norman Danner wrote:
Oh dear; hadn't realized this was an issue. We installed Twisted and Cyclone from source.
There is a debian package for twisted (python-twisted), but not for Cyclone.
I managed to install Cyclone with Sathya's apt-get/pip instructions. (I also had to apt-get install python-dev, or pip would fail; not that its error message was helpful, but Google was.)
So, I don't have a clear preference for or against Cyclone. On the one hand, getting rid of that dependency would be nice, in particular given that pip fails to handle basic problems such as a missing python-dev. But on the other hand, I'd like to get the current pyonionoo running without having to rewrite major parts of it first. How about we stick with Cyclone for the moment and postpone the decision of a possible rewrite towards a Twisted-only version?
Some more feedback after a _very_ quick look through the files:
- Why do we need something like scripts/cookie_secret.py?
We probably don't. We just jumped off of Sathya's first pass, and didn't play with the Cyclone details at all.
Sathya, what do you think about that file/feature? Can we get rid of it?
- Do the two handlers bandwidth.py and detail.py really need to list all
the fields in the respective documents? That seems hard to maintain when we make changes to the back-end. Ideally, the front-end would simply consider these files as valid JSON strings without parsing them at all.
No, those are old versions and will be completely replaced. We'll follow your approach for these handlers: use the query parameters to determine the fingerprints of routers that match the parameters, then just respond with the corresponding JSON files from the filesystem.
Hopefully we'll have an example of one of these that should enable one to plug in Sathya's new back-end without modifying the handler classes, just the database module.
Okay. Focusing on summary documents first then.
Thanks, Karsten