Cli fs operations refactoring

[azazel]

Each of the cli commands implements its own logic to do fs operations, both local-side and tahoe-side. I think that most of the command modules should be refactored to use common code. This is for many reasons among:

• remove duplication;
• code comprehensibility and normalization of behavior;
• spread availability of improvements found in latter commands (backup) to other commands;
• easier implementation of new features like non-ascii charset handling for filenames;
• better handling of logging?

The steps in order to do that (that I can see):

1. Recognize the needs of each command in term of fs manipulation;

2. Look at the best implementation available and eventually improve it;

3. maybe design a minimal layer that will help handling alias/path/cap based operations and other cases.

Other interesting points to discuss:

• See whether a replacement of httplib with twisted can help in parallelizing and increasing the bandwidth usage.

Let's discuss it!

[zooko]

azazel: Those are good reasons! We already have reasonable test coverage for most cli commands, so after doing this refactoring it should be possible to look at the per-line code coverage output and see that the tests are thoroughly exercising the code in the new refactored version.

I don't have time to do this myself (see my blog for notes about what I'm doing), but I would happily review and accept patches that do this (once the buildbot is green).

[davidsarah]

See whether a replacement of httplib with twisted can help in parallelizing and increasing the bandwidth usage.

That should probably be a separate ticket.