On Mon, 29 Jan 2007, Uwe Stöhr wrote:

 2. keep Joost's full installer but add Uwe's improvements. I prefer
    Joost's installer because it uses the same code as 1. This means
    less maintenance.

That's OK four you but I prefer to have something that is useful for most of the users and if this requires more work, I'll di this. And always maintained my code.

Hi Uwe. What is it that makes Joost's full/bundled installer to not be useful for you? I only tested them briefly, but on the surface they seem similar? (To me it seems like it would have most of the basic functionality you are interested in, leaving only a few things to add)

Joost Verburg schrieb:

 I will happily add all features of your installer that are missing in
 the official one, if the majority of the developers agrees with them.
 I will also help you if you prefer to merge the features yourself.

For that matter, I could help with getting the required functionality into the installer. It's bad if only one person understand the installer code. With a shared code base, we would ideally have three people with some understanding of it: Joost, Uwe and me.

(Personal comment: I just know I'm going to regret having worked on an installer for Windows... it might mean I'll have to do something similar at work in the future..)

I don't think its woth to invest lots of time again to be able to do this with the same code base. The major LyX part already has the same code base. You don't need to implement new features from my installer because my special features are designed to produce a bundled installer. So the easiest way is to leave your way the official installer and my way a LyX-distribution. I'm not reclaiming that the distribution will be official as I know that

Merging the code to have all features in one script wil cost lot of time, I fear that I then loose the overview and that it will be hard to maintain. Having two scripts for two different way doesn't harm. (Personylly I usually work this way also at work: Use a different script/file/machine/whatever for its special purpose.)

Let me see if I've understood your objections:
1. Your installer is already here and it's (almost?) finished
2. It is not worth the time to implement the additional features in
   Joost's code base. (Now?)
3. It is more work (now) to merge the code bases (which might delay 1.5?)
4. You are worried about losing overview/understanding of how the code
   works, which leads to difficulties in maintaining it.

However, this bit I don't get

        You don't need to implement new features from my installer because
        my special features are designed to produce a bundled installer.

Joost's code already has a version that is a bundled???

Besides argument #4 above, are there any reasons why Joost's code base can't be made to behave as desired?


In the long run I agree with several others that maintaining two code bases for the installers is a bad idea. However, how about the following plan:

Short term goal (applies to some date between LyX 1.5 and 1.6)
* There will be the following installers
  A. Minimal version from Joost (opt. downloads additional components)
  B. Complete version from Uwe

Long term goal (after LyX 1.5.0, before LyX 1.6):
* All installers are derived from the same code base (Joosts)
* There will be the following versions of the installers:
  A. Minimal version (optionally downloading additional components)
  B. Complete version - applications bundled including a LaTeX
If we feel it is desired, we could also in the future have a version:
  C. Semi-complete version - applications bundled except LaTeX

Consequences for the short term:
* Uwe ought to drop/remove his small installer from the public to avoid
  confusion.
* Uwe keeps and maintains his complete installer in SVN
* We don't advertise Joost's bundled installer to avoid confusion

There will only be a single official installer for LyX that primarily *only* installs LyX. In addition there is an unofficial installer that installs a LyX distribution. Anything more is a bonus.

Consequences for the long term:
We should implement the desired additional functionality from Uwe's complete installer into Joost's code base. This does however not have to be done before 1.5 is released. We should probably start discussing this in a separate thread. (Joost asked for this IIRC).

Going back to Uwe's objections:
Regarding 1-3: Yes, it is more work to merge the code bases. However, if it doesn't have to be done immediately and Joost as well as I are willing to do it (help in my case), I think that covers it. In addition, I and many others thinks it is worth the effort in the long run. If you feel you don't have the time or energy, you could take a more passive role. Or we could postpone when it is done. I think the important issue for many are that we are agreed it should be done sometime.

Regarding 4: Let's make it a requirement that the code is understandable and reasonably easy to mainain. Joost, do you see any problems with that? Are there any technical reasons why the could would be difficult to understand?

Thoughts?
/Christian

--
Christian Ridderström, +46-8-768 39 44               http://www.md.kth.se/~chr

Reply via email to