libc changes and freebsd-4 compatible binaries

Yury Tarasievich grog at grsu.by
Tue May 3 03:46:53 PDT 2005


Unfortunately, this begs for yet another reply, but I call for stopping it at 
that, okay?

On 3 May 2005 13:31, Jeroen Ruigrok/asmodai wrote:
> -On [20050503 12:22], Yury Tarasievich (grog at xxxxxxx) wrote:
> >So, as I see it, the bottom line is NO, after all.
>
> There's no way you can keep binaries working if you change the ABI.
> Or there might be if you employ all kinds of dirty tricks and whatnot.

I think, in fact, we are (or should be) talking userbase here. Each such jolt 
costs. And the product hasn't (yet?) any massive support behind it which 
would compensate for that.

> >Opera won't work (and this no mean thing), openoffice won't work (same),
> > and existing collections of freebsd-4 packages become useless. Compile
> > everything or die, eh? With all respect, but this is kicking people in
> > the teeth, just like what IBM did in their OS/2 times.
>
> Unfortunately such is the price of progress.
> There's a trade off you have to make between advancing the goals (finally
> wide character support, which was sorely needed) and keeping old stuff
> working all the time.
>
> And we already *knew* we cannot maintain eternal binary compatibility with
> FreeBSD.  

Good grief. Not two years have been passed after starting up, no userbase to 
talk about. Eternity??

> And there you see the exact reason why Theo de Raadt is such a 
> proponent of sources over binaries. ;)

Unfortunately, I've no time to discuss this extensively, nor is this the 
place, but, in a nutshell, the *real* exact reason for "sources over 
binaries" is desire to push more work/responsibility on to the user, and keep 
more fun for yourself. Don't I spill the same blood? :)
/the quotation may be incorrect, of course, but I hope you've got the 
point :)/






More information about the Kernel mailing list