[asdf-devel] [Sbcl-devel] Logical pathnames vs ASDF & SBCL

[Faré]

> Your reactions to bug reports are either null, or very discouraging.
>
Which bug did you report that I didn't address?
There are archives, please show me.

> I usually only get fixes or workarounds by people other than you.
>
I'm glad you're getting support from others. Unhappily,
you won't get any from me unless you communicate with me.

> Instead I repeatedly hear from you that either I am doing something wrong, or that what I'm trying to do is deprecated, or legacy, or old-style, or whatever.
>
I never said anything was "deprecated".
central-registry and logical pathnames are very much supported.
I did say legacy and old-style indeed.
Which means they continue to work as they always did.
Robert Goldman and I even took pains to maintain and extend
the documentation on how to use them, to include
common usage, known pitfalls and some advanced features.
While it's not a lot of documentation, it's still more than was before.

> Or you don't react at all (like with my _bug_ report that the documentation is misleading, which is the main reason why I switched from a working setup to a setup that doesn't work). Or like here, you react by saying that I'm not reporting problems properly, or not early enough, or whatever.
>
As far as I can tell, the reason your setup didn't work
is because the MCL support had bitrotten.
My excuse for that is that I never had a chance to use a Mac.
As far as I know, I fixed all the issues I am aware of
— based on binghe's bug reports.
I never had a clear bug report from you.

> I don't get such reactions from any other library/tool maintainer!
>
My apologies. I obviously fail to communicate with you.
But communication is a two-way thing.
Maybe you can come down to my level.

> You either support a feature, or you don't. Don't do this mixed bag of "I support it, but you're a loser if you use it."
>
I support central-registry and logical pathnames within my limits.
I didn't touch the legacy configuration system.
It's working, it's stable. You can keep using it as you always did before.
It takes priority over the new source-registry.
If you're using these features, though,
you're not a loser: you're an advanced user;
your curse is that you have to take responsibility for that,
because lesser people such as I are unable to do it.
You already know more about logical pathnames than I do;
I'm ready to learn from you everything you'll tell me
about how to do things right about them.

> The main reason why I'm having problems with ASDF 2.x is because it's alpha-quality software at best (you're still trying to figure out some very basic concepts, it seems to me), yet you don't even call it beta quality, but instead choose to call it 2.x, push it down everybody's throats, and turn a whole community into alpha testers who didn't volunteer to be guinea pigs in the first place.
>
I called it 2.x to make it clear that there had been significant changes
since the first 1.x series, and that while compatibility is our goal,
we warn that there might be breakage in a few
formerly underspecified or non-portable corner cases,
or on platforms we don't have access to.

I indeed am trying to "push it down everybody's throats",
in other words I aim to satisfy everyone enough that they will adopt it.
I apologize if you're experiencing trouble and encourage you
to pick whichever legacy version you're more satisfied with,
if you're experiencing issues with mine.
I certainly didn't install ASDF 2.x on your machine.
Why did you even try to upgrade?
If you needed a bug fix or new feature, then maybe it's not totally useless.
If you just wanted to test it, then I thank you for testing:
you were the very first to try ASDF 2 on RMCL.

> I have used ASDF 1.x and it's central registry approach for years in almost all CL implementations available, and I have never encountered any problems with it. I have also used logical pathnames to some limited extent years ago, and my experience with those weren't even remotely as bad as with ASDF 2.x. When using them recently, for a few weeks they actually made something work that ASDF 2.x claims to support, but actually didn't really, until ASDF screwed it up again.
>
My only claim of support is that I will fix bugs that are reported to me.
I can't fix bugs that are not reported.
If you have an unusual setup, you have to explain
what you are doing, what is breaking, and how it is breaking.
I'm not going to remotely hack into your computer at night
to find out whether things are wrong and why.
I unhappily can't run batteries of test
on an implementation I don't have access to.

That said, yes, it might be useful to write unit tests
for the support of logical pathnames by the source-registry.
Patches welcome.

> I'm not sure how to proceed from here. Communicating with you and reporting bugs doesn't seem to buy anything. If I had the choice I would just drop ASDF 2.x and go back to ASDF 1.x, but this doesn't seem a viable option. Or is it?
>
I don't see why you believe that ASDF 1.x isn't a viable option.
Certainly, if you believe that ASDF 2.x doesn't bring anything valuable,
you should stick to ASDF 1.x. Of course, I won't be able to provide
support for it, but that shouldn't be an issue for you,
since my support seems of little value to you.

Best regards,

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
Fascists divide in two categories: the fascists and the anti-fascists
	— Ennio Flaiano