Re: maintaining advanced power-user Scheme functions

[Phil Holmes]

> 2013/8/19 Phil Holmes address@hidden:

This will be my last post on this.


> > ----- Original Message ----- From: "Janek Warchol"
> > <address@hidden>
> >> 2013/8/19 Phil Holmes <address@hidden>:
> >> > However, my other opposition to git for this
> >> > purpose is that it's intended as a Version Control system, which is
> >> > completely unnecessary for the LSR - that's just a simple database of 
> >> >  >
> >> > text,
> >> > with no version control needed.
> >>
> >> Someone once said "if you have a project and it will take you more
> >> than an hour to recreate it from scratch, it should be under version
> >> control".
> >> It may not be apparent why this is so, but the experience confirms it.
> >
> > The LSR is not a project, it's a collection of text objects. A simple
> > database is the best place for a collection of text objects.
>
> Nonsense. I have a collection of lilypond scores (more than a
> hundred) in a git repository and it's working perfectly, and i have
> actually used version control features on that collection of scores.
> I.e. i did things that would be impossible to do without version
> control.
> Besides, look at Mutopia (it's a collection of text objects as well -
> i mean, the pieces are independent from each other). Or ask power
> users such as Nicolas, who has his scores under version control.
>
> What you said would be true if the snippets were so simple that they
> never change. But they do change.

Hardly ever.  I maintain the current LSR and the changes are so infrequent I 
often forget to check whether there are any.  This is part of the problem 
with your proposing changes - you admit you don't use the LSR, don't know 
how it works, don't know how to do simple updates, but you're proposing 
wholescale change.  Please try the current system before trying to make 
large changes to it.

> >> > If you would actually try the LSR, you'll see it's equally quick to >
> >> > enter
> >> > snippets - each takes a few seconds or a minute at most. All that you 
> >> >  >
> >> > need
> >> > to do is type the snippet (or paste it) and a description - you can't 
> >> > do
> >> > _less_ with any other system.
> >>
> >> Can i do it offline?
> >
> > You can't update any remote system offline - git or the LSR.
>
> but with git i can do everything that matters offline - add new stuff,
> fix stuff, search old versions, etc. Then i just press a button and
> it does the rest.
> Imagine that i'm working on a big project w/o internet access (because
> i'm on holidays) and i create a few snipets. Do you think i'll
> remember them when i get back online? Definitely no! I will only
> care to add them to the repository when i'm actually working on the
> stuff, not a week later.

You could consider a text editor to create the snippets offline, giving them 
sensible names in a directory you can remember, and pasting them when you're 
online.

> > What you've described there is a set of odd bits of code that are badly
> > structured and with no clear idea of what they do ("chaos here" is not 
> > good
> > documentation practice). Putting them in a git repo would not make them 
> > any
> > more suitable for our documentation system than putting them in the LSR. 
> > For
> > your own use - fine.
>
> What's funny is that this description fits out issue tracker quite
> accurately. It's a set of odd bits of code, usually with no clear
> idea what they do (i mean, usually we can see that the output is
> wrong, but we don't know what actually happens inside - if we knew,
> many more bugs would be fixed instantly after submitting). And they
> are not structured in any way.
> Yet we keep them, because they contain some valuable information and
> it's better to have some information rather than none at all.
> Moreover, we have a policy to add issues to the tracker quite freely,
> i mean - when someone posts something wrong, we don't analyze it
> trying to find connections with other issues, but we just throw it in.
> It may be marked as duplicate later.
>
> Yet this system seems to work quite well.

Because it's designed to work as a way of remembering odd bits of 
information, not as a way of documenting how LilyPond works.  The LSR is 
part of the LilyPond documentation system that users can access and read. 
It must be an accurate set of information and that's how it currently works.

> > The LSR is intended to allow non-power users the ability to contribute
> > lilypond code to other users and potentially our documentation. As such, 
> > it
> > performs a useful service. I cannot imagine why you want to go to all 
> > the
> > work of reimplmenting something that fundamentally works using completly
> > different technology. It would be a massive waste of time.
>
> Well, as for "completely different technology", that's not necessarily
> what i propose. Using git as the tool to manage the actual *content*
> doesn't have to mean changing thee tools that operate on that content.
> Why do i want to do this? Because the current system is limited: we
> don't have support for more than one lily version, upgrading is
> inconvenient, and there are no convenient ways to add some valuable
> stuff there.

Fine.  If you want to waste your time, feel free to propose to change a 
perfectly adequate database for a completely different technology.  You will 
need to start by contacting the maintainer of the LSR, who can be difficult 
to get hold of.  That's the real problem...

--
Phil Holmes