> Gavin Sinclair wrote:
>> Quick update: I ironed out the marshal problem. I needed to read the
>> file in binary mode. D'oh!
>>
>> My attention now will be generating more class descriptions and
>> cleaning up the existing ones a little.
>
> Even with marshal working, is it still the format we want to use?
> If we're moving away from ri-reuse, then perhaps we should use a disk
> format that is human-readable, easy to search, easy to cross reference,
> and easy to apply the usual CRUD (create, read, update, and delete)
> operations.
It's still eminently suitable for ri/rj, as you want quick answers to
known queries, so you don't want to load much. I think Dave's vision
requires a database, but that's a lot of effort, which is the antithesis
of rj.
I like human-readable data, and YAML is great for that. YAML and marshal
share the very useful property that they are serialization formats. Any
other format requires us to write and test an adapter, which I don't like
the sound of.
The mark against YAML at the moment is that it's immature. It's good, but
bugs are still being found. One incedent this year notwithstanding,
marshal is reliable.
I'd be interested to know what features you can think of that would
benefit from CRUD abilities.
> I suggested using DAB's plain-text database because it may make it much
> easier to get interesting, flexible behavior from rj.
>
> After a quick look at the README[0] I'm not entirely sure this is the
> best choice, but I would think there exists somethng along these lines
> we can use.
Cheers,
Gavin
|
