On 06.04.2018 11:18, Jean-Marc Lasgouttes wrote:
Le 06/04/2018 à 00:40, Uwe Stöhr a écrit :
Am 05.04.2018 um 04:47 schrieb Scott Kostyshak:
In my opinion it is that important, because updating LyX could break
something else on a user's computer.
No! Why do you claim this again? Don't mix potential bugs in a LaTeX
package on CTAN with LyX. With this argumentation every package update
could introduce bugs. LyX is not responsible for bugs on CTAN.
Aren't we discussing bugs in miktex here? CTAN is a different beast,
although living on the bleeding edge is probably a bad decision for
"average" users.
Yet another attempt to clarify a bit (maybe just for me):
I was hoping the discussion moved past the point discussed above. If
not, we should first try to answer the following question:
(Q1) Does installing a new version of (or update) MiKTeX can break the
compilation of LaTeX documents that were compiled fine before (because
it can introduce bugs that an already installed previous version did not
have)?
It seemed to me that there is agreement on this. Please correct me if I
am wrong.
Only because (Q1) can be affirmed, the second question became important:
(Q2) During installation, should there be a question asking the user
whether to update (and, alternatively), cancel the installation?
At least two principled approaches have been suggested:
First, what I call
*the consent approach*: Yes to Q2, because LyX should not do without
consent what can break documents that compiled fine before.
If I see it correctly, this approach was applied in previous version
where the user could decline an update of MiKTeX (though it was
recommended to update).
Second, what I call
*the silent approach*: No to Q2, because the LyX installation should
automatically do what is best for (most?) unexperienced users and a
dialog could confuse them.
Now, if I see it correctly, the matter to decide the issue is one of
importance. It's a value judgment. And there is no dominant approach
that is better in all respects. Notice also that one can find both what
is best for unexperienced users *and* not to break documents that
compiled fine before valuable. It is just that one comes out on one or
the other side because of an overall judgment, a favorable combination
of these values.
In so far, I think no side is being stupid by overlooking an totally
obviously better approach. If this is correct, I think it is very
important that everyone sees this since it helps both to understand the
other side and not to feel like others think that one is just being stupid.
Some have tried to lessen the strength of the reasons in favor of the
silent approach by making the dialog easier to understand. If this would
be fully successful, then the consent approach could come closer to
being dominant in all respects. (For example, Scott is following this line.)
On the other hand, it might be possible to lessen the reason in favor of
the consent approach, for example, if MiKTex updates almost never break
a LaTeX system. This might be doubted given the reason the current
problem was arrived at.
I see the following ways forward:
1. Further trying to discuss the basic values at stake and see whether
one approach comes out - even if not in all respects - overall
favorable. (For example, Uwe and Jean-Marc are following this approach
by discussing analogue cases, like starting a car.)
2. Act on a set of basic principles (like a charter) the LyX development
should follow, if there is one. Maybe there is a rule like: Never do
anything without consent that could break the LaTeX compilation, or
always do what is best for the unexperienced user.
3. Use some collective method of decision making common in democracies,
like majority or two-thirds majority (among developers). And then just
act on it together. Obviously, decision making using vetoes would not
work at the moment.
All other alternatives I can think of are kind of sad... but maybe I
overlooked something.
Best,
Daniel
