Date: Thu, 25 Mar 1999 00:07:37 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: brett@peloton.physics.montana.edu (Brett Taylor) Cc: brett@lariat.org, jkh@zippy.cdrom.com, advocacy@FreeBSD.ORG Subject: Re: FreeBSD emulation for linux Message-ID: <199903250007.RAA05081@usr09.primenet.com> In-Reply-To: <Pine.BSF.4.05.9903222133500.1749-100000@peloton.physics.montana.edu> from "Brett Taylor" at Mar 22, 99 10:07:05 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > I'm beginning to think it's not worth even trying to talk sense to you
> > folks regarding PR, much less contribute time or code.
>
> You've suggested the following things that I remember recently:
>
> - that the Linux emulator has been bad for FreeBSD
>
> This is contrary to every person I've hooked on FreeBSD. They appreciate
> the fact that its stable AND that they can run the preponderance of Linux
> apps that are available.
I'm actually the origin of this idea. Well, not actually. Ashton-Tate
is the true origin.
I'm dating myself by saying this, but....
Back in the days of dBase III, anyone who wanted to sell dBase III
based programs had to not only sell their program to their customer,
but they also had to sell them dBase III.
It doesn't take a rocket scientist to see why Clipper made such a big
splash: by removing the dBase III runtime requirement, it halved the
amount of selling you had to do.
I think that Linux emulation is bounded by two arguments:
1) It's clearly necessary, in that it allows FreeBSD to woo
Linux users to FreeBSD, and to bolster sagging native
applications statistics with Linux statistics.
2) It's clearly damaging to the availability of native
FreeBSD software, since it (effectively) pays developers
a positive differential for *only* developeing for Linux,
rather than creating Linux AND FreeBSD native binaries.
At this point someone will yell "CDE! CDE!", and point out that,
because FreeBSD users didn't want to buy a particular niche native
binary from a vendor who had been sold a rope, that there therefore
must be no market whatsoever for native FreeBSD binaries.
The problem with this particular instance is that FreeBSD "Osborne'd"
the native vendor by promising a "soon-to-be-released-new-version",
and the vendor was over-sold on the potential market, and then
under-delivered.
There is a market for FreeBSD native binaries. This can be stated
as fact, given that people buy binaries (including Linux binaries)
to run on FreeBSD. But how do you size this market?
First, you have to subtract out all of the people running the Linux
software under emulation because they simply couldn't wait. Then
you have to take the support differential, and compare it to the
remaining number of potential sales that a native binary would buy
you.
You may damn the source, but Brett's OS/2 analogy is apt.
I personally think that a FreeBSD emulation layer for Linux, and
other OS's, is important for FreeBSD, from a marketing standpoint.
From a technical standpoint, Solaris and SCO are implementing Linux
emulation as well. The Linux ABI looks to be gaining momentum as
the cross-x86 UNIX ABI. This is technically very bad. Consequences
range from loss of editorial control of the FreeBSD ABI to what pass
as architects in the Linux camps(*) all the way through rendering
FreeBSD meaningless to everyone but people with direct control over
its direction -- practically a branch of Linux.
It's important that FreeBSD define its long term goals, such that
short term decisions, such as whether or not a FreeBSD ABI for Linux
should receive project attention or effort, can be held up against
those long term goals.
FreeBSD needs to devise a vision that can be used as a litmus test,
to keep it from ratholing down discussions on whether proposed work
has merit or not.
(*) I am not criticizing Linux coders here; being agile has its
merits. But being agile in your ABI definition is frankly a
nightmare for the poor application vendors chained to the
back of your pickup truck like a forgotton hound.
Terry Lambert
terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.
>
> - that the ports tree should support every version of FreeBSD
>
> I asked if you'd even approached Satoshi or Mike Smith about your idea,
> since they are working on the new package system. You haven't answered
> that yet. Do you have anything in mind other than this vague, if not
> noble, idea, like how possibly to implement it? Do you _really_
> understand how the current ports system works? That's step one. Talking
> to Mike and/or Satoshi is step 2. And not just "I'd like to have X
> happen," wo/ any idea how to implement it.
>
> - that a FreeBSD emulator should be created for Linux in the hopes
> that this would make people write directly for FreeBSD
>
> You are the _only_ one I've heard support this. Not one person to my
> recall has suggested this is a workable idea other than you. Sure it'd be
> nice if every app was built to run natively on FreeBSD, but it's not
> reality nor is that likely to change just because an emulator exists.
> Companies will say "sure there's this FreeBSD emulator for Linux, but they
> have roughly 1/6th the number of installations _and_ can emulate Linux.
> Let's just write for Linux." Linux people will certainly not drop writing
> for the Linux API just because there's suddenly a FreeBSD emulator and if
> you really believe that ...
>
>
> The problem Brett is that all of the things you have suggested/proposed
> have _no_ backing from anyone but you (for whatever reasons), I believe
> because they sound mostly like ideas wo/ any real plan on how to do it.
> I'd love to see a ports mechanism that supports all the FreeBSD versions
> (at least back to 2.2.8), but I doubt it'll ever happen. There just
> aren't people to work on it. You keep saying you'll write code but I'm
> not even convinced you have a clear idea of how you even want to start
> this ports project you suggested.
>
> Even if you couldn't do this, you could, in a relatively easy manner (at
> least compared to this nebulous porting mechanism) try to just maintain a
> tree of ports that are up-to-date for 2.2.8. You'd never be able to do it
> alone, but I'm sure if you asked nicely and had some, say 10, people who
> were really dedicated it could be done w/ a lot of work.
>
> As it is I haven't even seen you attempt to fix or maintain one port. I
> suggested doing libgtop. You refused that because "it's GPL." I got news
> for you - people want to run Gnome (go see -questions or -ports for an
> idea of how many people are trying to get them to compile) and Gnome needs
> glibtop.
>
> If it's still such a personal affront to you to maintain/fix a port that
> is GPL'ed go find a port that you use that has:
>
> MAINTAINER= ports@freebsd.org
>
> that isn't GPL'ed and maintain it. As I told you before, there are lots,
> 145 in my tree (wo/ any of the foreign language ports), of ports which
> need maintainers. Maintaining a single port would be a great way for you
> to get a good introductory knowledge of how the ports system works and how
> you might be able to implement your idea. Heck Bill Fenner sends out a
> "homeless ports that need help" list once a week. Pick one from the list
> and fix it. Often times it just needs to have the distfile updated. But
> again, you haven't even done that.
>
> Since you've been posting regularly to -advocacy, the only thing I've seen
> from you is grandiose ideas and no clear idea how to implement them or
> even willingness to do any work as simple as it may be (such as
> maintaining a port or fixing it so it will work for 2.2.8).
>
> Jordan hit it on the head when he said you just seem to shoot from the
> hip, seeing what you hit.
>
> Brett Taylor
> ***********************************************************
> Brett Taylor brett@peloton.physics.montana.edu *
> brett@daemonnews.org *
> *
> http://www.daemonnews.org/ *
> ***********************************************************
>
>
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-advocacy" in the body of the message
>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-advocacy" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903250007.RAA05081>