Also, there's a thing that I've been trying to figure out
since I first implemented the "prep custom" assistant:
Should it precisely remember the dependencies that it was created with or should it
dynamically load the dependencies from the creator assistant? Both have some pros and
cons:
- precisely remember dependencies:
Pros:
- You created the project in a certain way with a certain dependency set. If the
creator assistant changes, your project stays the same => you want the same dependency
set that you started with.
- You can always erase the original dependencies completely and rewrite them as you
like.
Cons:
- If you create a system on e.g. F20 and the dependency names change in e.g. F22, your
.devassistant will still have the old names and won't work well.
- dynamically load dependencies - the opposite of above
So it's also a sort of philosophical question - we started with "dynamically
load dependencies", but I think we should now move to "precisely remember
dependencies" for the above reasons.
I see what you mean. I had (a dangerous) idea, but more I think about
it, I think it will be very complicated.
The idea was (and I don't think it's a good one):
* Let's implement dependency guessers for various types of dependency
listing. As start, take requirements.txt for Python.
But that would probably only work with PyPi and not with RPM and stuff
like that anyway, and we would have to do that for lots and lots of
languages and frameworks.
So let's not go there :D
(I will send the e-mail anyway, so you and others can see the idea as
well, but don't take it to seriously.)
--
Miro Hrončok
--
Phone: +420777974800
IRC: mhroncok