Hi Harm, I see your input, but I’m comparing this with how it looked like 5 years ago: I could not even compile my own songs anymore with Guile 2.x installed.
First for the most important part I see in your mail: > And we still may be dropped from major distros or lilypond with guile2 > is offered, which may give us bad reputation. This choice is much, much better than before. The choice before was "I’m losing Lilypond!" (and not much of a choice). Now the choice is: "Lilypond might be slower" Whether being slower actually gives Lilypond a bad reputation is mostly unclear. You show me a factor of 2 in speed for small documents. That means: It works. And factor 2 is actually much better than I thought. The non-stability for your larger score is something to fix, but it is something which can actually be addressed systematically and no longer a roadblock users cannot cross without losing all ability to compile their music. And it is something which hits people who might actually have the skills to fix it (those who embedded scheme in the lily documents). With this I do not want to say that everything is rosy. There for sure are challenges. What I want to say is that there is a world of difference between the situation now and the situation a few years ago. And there is no longer reason for gloom: The challenges now look solvable in principle and an praxis. It’s still unnerving that additional work comes, but now there’s a path and people can go for it and optimize Lilypond for Guile 2.x, as it was optimized for Guile 1.8 before. Keep in mind what Andy Wingo said at FOSDEM 2016: With Guile 2.x we need to update our performance model: https://archive.fosdem.org/2016/schedule/event/guilevm/ Best wishes, Arne -- Unpolitisch sein heißt politisch sein ohne es zu merken
signature.asc
Description: PGP signature
