New domain!

New domain!

[Rich Felker]

Hi all,

As you can see from the mail headers, we have a new domain to use:
libc.org. This is thanks to Kurt Maier (you rock!) who managed to get
it and donated it for use by musl.

There's been some limited discussion so far on IRC of how to best use
it, but I think it makes more sense to move that over to the list. A
summary of the ideas so far (not all of them are necessarily going to
happen, of course):

- Using musl.libc.org subdomain as either the main website or a
  redirect to the main website on the existing domain.

- Or just being pretentious like the kernel folks (think kernel.org)
  and having libc.org be the musl site. ;-)

- Revamped libc comparison with more rigorous results, more libcs
  compared. (This is an existing project idea, but new domain is a
  good place to put it.)

- Information on standards, platform ABIs, etc.

- Browsable 3p man pages.

- Why ppl should care about libc, standards and interoperability, why
  existence of multiple implementations of any important library is a
  good thing, etc.

Further ideas or discussion of the existing ones is welcome here.

Thanks again to Kurt Maier!
Cheers.

Rich

Re: New domain!

[Luca Barbato]

On 16/04/14 17:52, Rich Felker wrote:
> Hi all,
> 
> As you can see from the mail headers, we have a new domain to use:
> libc.org. This is thanks to Kurt Maier (you rock!) who managed to get
> it and donated it for use by musl.
> 
> There's been some limited discussion so far on IRC of how to best use
> it, but I think it makes more sense to move that over to the list. A
> summary of the ideas so far (not all of them are necessarily going to
> happen, of course):
> 
> - Using musl.libc.org subdomain as either the main website or a
>   redirect to the main website on the existing domain.

Do you want to host other libc in other subdomains?

> - Or just being pretentious like the kernel folks (think kernel.org)
>   and having libc.org be the musl site. ;-)

Would be interesting =)

git.libc.org is neat anyway

> - Revamped libc comparison with more rigorous results, more libcs
>   compared. (This is an existing project idea, but new domain is a
>   good place to put it.)

Would be good to have no matter where.

> - Information on standards, platform ABIs, etc.

Nice

> - Browsable 3p man pages.

Would be much welcome.

lu

Re: New domain!

[Rich Felker]

On Wed, Apr 16, 2014 at 06:27:03PM +0200, Luca Barbato wrote:
> On 16/04/14 17:52, Rich Felker wrote:
> > Hi all,
> > 
> > As you can see from the mail headers, we have a new domain to use:
> > libc.org. This is thanks to Kurt Maier (you rock!) who managed to get
> > it and donated it for use by musl.
> > 
> > There's been some limited discussion so far on IRC of how to best use
> > it, but I think it makes more sense to move that over to the list. A
> > summary of the ideas so far (not all of them are necessarily going to
> > happen, of course):
> > 
> > - Using musl.libc.org subdomain as either the main website or a
> >   redirect to the main website on the existing domain.
> 
> Do you want to host other libc in other subdomains?

No, not particularly. But it reads nice that way anyway, and I figure
since we have it we should use it.

> > - Or just being pretentious like the kernel folks (think kernel.org)
> >   and having libc.org be the musl site. ;-)
> 
> Would be interesting =)
> 
> git.libc.org is neat anyway

Yes, this (git.) is definitely a possibility even if the main
second-level domain is not used for musl website.

> > - Revamped libc comparison with more rigorous results, more libcs
> >   compared. (This is an existing project idea, but new domain is a
> >   good place to put it.)
> 
> Would be good to have no matter where.

Yes. This is something I'd like help on, and it doesn't really need a
lot of expertise. Basically the part that could use some work would be
a matter of improving the actual measurements (e.g. adding min/avg
runtime for certain interesting calls, and throughput in MB/sec for
various usage patterns, etc. where it makes sense, rather than just
arbitrary runtimes with no context) and automating the process of
testing sizes and speeds.

Rich

Re: New domain!

[Laurent Bercot]

On 16/04/2014 16:52, Rich Felker wrote:
> As you can see from the mail headers, we have a new domain to use:
> libc.org. This is thanks to Kurt Maier (you rock!) who managed to get
> it and donated it for use by musl.

  Now THAT is awesome.


> - Using musl.libc.org subdomain as either the main website or a
>    redirect to the main website on the existing domain.

  Sounds like a given.
  I find musl-libc.org redirecting to musl.libc.org more elegant
than the other way around, but YMMV.


> - Or just being pretentious like the kernel folks (think kernel.org)
>    and having libc.org be the musl site. ;-)

  So tempting. What can we do to tempt you even more ?


> - Revamped libc comparison with more rigorous results, more libcs
>    compared. (This is an existing project idea, but new domain is a
>    good place to put it.)

  That sounds good too, but don't you think people will question the
objectivity of libc.org comparisons if the domain belongs to musl
authors ? The current comparison on musl-libc.org is *expected* to
be biased, even if it is not.
  Sigh. Managing appearances is hard.


> - Information on standards, platform ABIs, etc.
> - Browsable 3p man pages.
> - Why ppl should care about libc, standards and interoperability, why
>    existence of multiple implementations of any important library is a
>    good thing, etc.

  I would love all of those. I might even contribute if I get some
free time by way of divine intervention.


> Further ideas or discussion of the existing ones is welcome here.

  e-mail addresses, of course !
  Please can I have ska@libc.org, please please pretty please ?
I'll draw you a kitty. Or contribute something if I can find something
I'm more knowledgeable about than you guys.

-- 
  Laurent

Re: New domain!

[Rich Felker]

On Wed, Apr 16, 2014 at 09:10:32PM +0100, Laurent Bercot wrote:
> >- Using musl.libc.org subdomain as either the main website or a
> >   redirect to the main website on the existing domain.
> 
>  Sounds like a given.
>  I find musl-libc.org redirecting to musl.libc.org more elegant
> than the other way around, but YMMV.

I agree, but that takes a little bit of work transitioning links and
getting search engines to recognize the move, and it's been a long
time since I've done such a move so I have to figure out how it works
again.

> >- Or just being pretentious like the kernel folks (think kernel.org)
> >   and having libc.org be the musl site. ;-)
> 
>  So tempting. What can we do to tempt you even more ?

I think it's more just a matter of the practical pros and cons, and
one issue is establishing name recognition. musl.libc.org establishes
name recognition for musl and plain libc.org or www.libc.org doesn't.
So I'm leaning towards an approach that uses them in conjunction with
musl.libc.org being the musl site and libc.org being something that's
not misrepresented as fully independent of musl, but that's also more
general.

> >- Revamped libc comparison with more rigorous results, more libcs
> >   compared. (This is an existing project idea, but new domain is a
> >   good place to put it.)
> 
>  That sounds good too, but don't you think people will question the
> objectivity of libc.org comparisons if the domain belongs to musl
> authors ? The current comparison on musl-libc.org is *expected* to
> be biased, even if it is not.
>  Sigh. Managing appearances is hard.

I think it's already stated that the main window for bias is in the
choices of what aspects are important to compare. (Note: This same
principle applies to basically everything: news, surveys, academic
research, fiction, etc.; the choice of what story to tell is the
biggest bias of all.) Once that's understood, I don't think there's a
lot of remaining question of bias. Most of the items are either
factual and easily verifiable/falsifiable, or quantitative
measurements.

For the revamp though I think it would be nice to put some more effort
into reducing bias in the choice of what to compare. I'm not sure if
it would make sense to ask glibc and/or uclibc to suggest additions
but that would at least be one thought for how to do it.

> >- Information on standards, platform ABIs, etc.
> >- Browsable 3p man pages.
> >- Why ppl should care about libc, standards and interoperability, why
> >   existence of multiple implementations of any important library is a
> >   good thing, etc.
> 
>  I would love all of those. I might even contribute if I get some
> free time by way of divine intervention.

Another one to add: why C is important and has not just modern
relevance but has concrete positive aspects that other languages fail
to duplicate that makes it basically the only language for systems
programming, etc.

> >Further ideas or discussion of the existing ones is welcome here.
> 
>  e-mail addresses, of course !
>  Please can I have ska@libc.org, please please pretty please ?
> I'll draw you a kitty. Or contribute something if I can find something
> I'm more knowledgeable about than you guys.

You're the second to request this. It depends on some mail server
setup and thinking through which systems should be responsible for
having mail going thru them, but it's something I hope to add before
too long.

Rich

Re: New domain!

[Laurent Bercot]

On 17/04/2014 05:18, Rich Felker wrote:

> I agree, but that takes a little bit of work transitioning links and
> getting search engines to recognize the move, and it's been a long
> time since I've done such a move so I have to figure out how it works
> again.

  Transitioning links is a perl one-liner, and putting in a redirection
should also be a one-liner in the web server configuration.


> So I'm leaning towards an approach that uses them in conjunction with
> musl.libc.org being the musl site and libc.org being something that's
> not misrepresented as fully independent of musl, but that's also more
> general.

  That sounds like the best way to go about it.


> For the revamp though I think it would be nice to put some more effort
> into reducing bias in the choice of what to compare. I'm not sure if
> it would make sense to ask glibc and/or uclibc to suggest additions
> but that would at least be one thought for how to do it.

  It would at least make sense to ask. If only to read the hilarious
subsequent bikeshedding discussion on the glibc mailing-list. :)


> Another one to add: why C is important and has not just modern
> relevance but has concrete positive aspects that other languages fail
> to duplicate that makes it basically the only language for systems
> programming, etc.

  I can help with that.


> You're the second to request this. It depends on some mail server
> setup and thinking through which systems should be responsible for
> having mail going thru them, but it's something I hope to add before
> too long.

  I wasn't thinking about mail hosting, but only about mail aliases.
I can help with that too (definitely aliases, and even hosting to some
extent).

-- 
  Laurent

Re: New domain!

[Rich Felker]

On Thu, Apr 17, 2014 at 09:36:54AM +0100, Laurent Bercot wrote:
> On 17/04/2014 05:18, Rich Felker wrote:
> 
> >I agree, but that takes a little bit of work transitioning links and
> >getting search engines to recognize the move, and it's been a long
> >time since I've done such a move so I have to figure out how it works
> >again.
> 
>  Transitioning links is a perl one-liner, and putting in a redirection
> should also be a one-liner in the web server configuration.

I meant links from other third-party sites and especially getting
search engines to return the new canonical site. If you do it wrong,
you'll drop in the search results; right now we're #2 for musl which
is pretty good considering #1 is musl.com and #3 is musl.net. BTW
having libc.org might eventually put musl at #1 for "libc"; right now
we're #7. :)

> >So I'm leaning towards an approach that uses them in conjunction with
> >musl.libc.org being the musl site and libc.org being something that's
> >not misrepresented as fully independent of musl, but that's also more
> >general.
> 
>  That sounds like the best way to go about it.

That seems to be the general consensus so far.

> >For the revamp though I think it would be nice to put some more effort
> >into reducing bias in the choice of what to compare. I'm not sure if
> >it would make sense to ask glibc and/or uclibc to suggest additions
> >but that would at least be one thought for how to do it.
> 
>  It would at least make sense to ask. If only to read the hilarious
> subsequent bikeshedding discussion on the glibc mailing-list. :)

OK. A couple ideas for things to add that come to mind:

- Compatibility with major proprietary dynamic-linked software (glibc
  full, musl partial, others none)

- Sizes with minimal configuration (would show how small you can get
  uclibc if you cripple it as much as possible)

- Time to compile (probably reflects most positively on musl)

- Time to compile a given medium-size app against the libc (measures
  header bloat cost) but it's hard to factor out toolchain differences
  that might not be caused by libc, and even harder if you also want
  to measure configure time

> >You're the second to request this. It depends on some mail server
> >setup and thinking through which systems should be responsible for
> >having mail going thru them, but it's something I hope to add before
> >too long.
> 
>  I wasn't thinking about mail hosting, but only about mail aliases.
> I can help with that too (definitely aliases, and even hosting to some
> extent).

Well the current setup has the mx going to a system that's not setup
for forwarding mail back out at all (only local delivery). I would
probably move it to the musl vps to add more features, so hosting
isn't so much of a problem, but setup is. Securing a mail system that
can end up sending outgoing mail is a big deal and really sucks if you
get it wrong (spam liability).

Rich

Re: New domain!

[M Farkas-Dyck]

On 18/04/2014, Rich Felker <dalias@libc.org> wrote:
> I think it's more just a matter of the practical pros and cons, and
> one issue is establishing name recognition. musl.libc.org establishes
> name recognition for musl and plain libc.org or www.libc.org doesn't.

Like kernel.org. I think they at least ought to have registered
linux.org too, but meh.

> So I'm leaning towards an approach that uses them in conjunction with
> musl.libc.org being the musl site and libc.org being something that's
> not misrepresented as fully independent of musl, but that's also more
> general.

This makes sense.

> Securing a mail system that
> can end up sending outgoing mail is a big deal and really sucks if you
> get it wrong (spam liability).

I doubt whether it's worth the bother. That said, if you think so, I'd
like <strake@libc.org> ☺

Re: New domain!

[u-igbb]

On Wed, Apr 16, 2014 at 11:52:11AM -0400, Rich Felker wrote:
> As you can see from the mail headers, we have a new domain to use:
> libc.org. This is thanks to Kurt Maier (you rock!) who managed to get
> it and donated it for use by musl.

That's a terrific news.

> - Using musl.libc.org subdomain as either the main website or a

Looks reasonable.

>   redirect to the main website on the existing domain.

No need imho. Rather the other way around for some time.

> - Or just being pretentious like the kernel folks (think kernel.org)
>   and having libc.org be the musl site. ;-)

I happen to dislike kernel.org for this very reason. :)

> - Revamped libc comparison with more rigorous results, more libcs
 [skipped more ideas]

Everything looks good, given that there is somebody to actually do that.

> Further ideas or discussion of the existing ones is welcome here.

Given the generality of the "libc.org" name it should possibly have place
for minimal information about non-Linux libcs as well, like a "link page"?

<xxxx>.libc.org
showing the mentioned "link page" or possibly a separate page per
library with some short info and a link to klick on,
for xxxx in uclibc glibc and so on?

> Thanks again to Kurt Maier!

 +1

Rune

Re: New domain!

[Anthony G. Basile]

On 04/16/2014 11:52 AM, Rich Felker wrote:
> Hi all,
>
>
> - Information on standards, platform ABIs, etc.
>
> - Browsable 3p man pages.
>
> - Why ppl should care about libc, standards and interoperability, why
>    existence of multiple implementations of any important library is a
>    good thing, etc.

Bikeshedding time!

I would like to see the above three emphasized.  Your libc is only 
slightly less important than your kernel, but this is certainly not the 
case with chatter in the wild.  It would be nice to have a site which 
spells out what libc is or should be.  And I would like to be able to 
use the site to learn stuff.

I would not make libc.org = musl but use the above to justify/explain 
musl for the same reason that kernel.org = linux kernel is pretentious.


-- 
Anthony G. Basile, Ph. D.
Chair of Information Technology
D'Youville College
Buffalo, NY 14201
(716) 829-8197