Glenn, what really angers me is that the requests for grep/od/tr/xargs
integration into libcmd have been turned down on the basis of 'too
much risk', but in the same time the x=$() doomsday bug crept in and
wasn't handled quickly despite a LONG queue of warnings and that a
"stable" ksh93u+ was released despite the bug was not corrected.

On 13 June 2012 15:43, Glenn Fowler <[email protected]> wrote:
>
> [ I added ast-users back to the forwarding list ]
>
> On Wed, 13 Jun 2012 13:04:33 +0200 Lionel Cons wrote:
>> On 13 June 2012 12:39, Lionel Cons <[email protected]> wrote:
>> > On 13 June 2012 06:35, Glenn Fowler <[email protected]> wrote:
>> >> the AT&T Software Technology ast beta 2012-06-12 source release
>> >> has been posted to the download site
>> >>        http://www.research.att.com/sw/download/beta/
>> >> the package names and md5 checksums are
>> >>            INIT  534d4fa288fdcf79141b1ec2e71bee8d
>> >>        ast-base  6a08c19d22027433b0808279e6638097
>> >>        ast-open  097c6578696efc82143dbf7fd15b7a83
>> >>         ast-ksh  4e2651f7ca60d73636bbe02576ca50d7
>> >> the md5 sums should match the ones listed on the download page
>> >> if not then don't download
>> >
>> > grep wasn't moved to ast-ksh, right?
>
>> Glenn, David, what exactly is the technical problem which prevents
>> grep, od, tr and xargs from being moved to ast-ksh/libcmd? I don't
>> understand the technical motivation of not doing this.
>
> we've stated our case
> I'll restate one more time
> I guess we have a larger than average (internal not-open-source) user base to 
> deal with
> a lot of ksh usage within that user base is on critical paths
> any disruptions obviously cause delays, but it can also damage ksh's rep
> and make it harder for us to convince projects to use our stuff
> many times ksh and the standard (1) commands are a foot in the door
> to other software that we first peddle internally, and then externally
>
> even the seemingly most simple change can have unforseen consequences
> adding a builtin like grep is simple technically
> but all of the consequences are not known
> operating inside the ksh address space requires a lot of coding discipline
> we don't yet know all of the details of that discipline
> or its manifestation on different platforms
>
> sure it was exercised some on linux and solaris
> but we have hp, bsd, aix, mvs, cygwin, windows instantiations that haven't
> gotten the same treatment
>
> so please set aside the mostly valid arguments for moving FOO to libcmd
> (convenience, efficiency, no approval needed, etc.) and let -lcmdtst
> burn in for a while -- let it be thouroughly *beta* tested -- let it be tested
> in a way that if a completely catostrophic flaw were found users would simply
> edit a non-default setting in a .paths file to work around it -- let it
> be tested to the point where we have a collection of regression tests that
> we feel cover all of the problem areas -- in particular for grep we need
> some pty(1) tests to cover all of the interactive interrupt-laden scenarios
>
> also, let the -lcmdtst paradigm itself be thouroughly tested
> after all, this will be the mechanism for most orgainzation to add their own 
> builtins
>
> btw, "thouroughly tested" is not a voting process, ``"worked for me" N times 
> => its good''
> for ast it implies a good faith effort to supply regression tests that hook 
> into the build system
> which automates the testing grunt work (this was learnt over time -- in 
> general in ast the
> parts that have the most tests are the newer parts -- tests for older parts 
> happen when they
> contribute to bugs or are otherwise updated)
> its fairly easy to get a base set of tests to work the first time
> its much harder to code software that has no regressions as the environment 
> around it evolves
>
> please suffer with what may seem to be a non-ideal interface for the better 
> good
> eventually, if your favorite command migrates from -lcmdtst to -lcmd,
> *you won't have to do anything* to take advantage of it it
>
> I won't explain this position again for the rest of the year
>
> on that note
>
> how about chanelling this -lcmd energy to working on a short list of 
> candiates for -lcmdtst
> my first one is ls(1) for practical reasons
> it seem that windows8 "power shell" (I think its cmd.exe) has dir builtin, 
> and apparently
> no dir.exe, so debugging the bootstrap uwin env from ksh gets tedious
>
> _______________________________________________
> ast-users mailing list
> [email protected]
> https://mailman.research.att.com/mailman/listinfo/ast-users
_______________________________________________
ast-users mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-users

Reply via email to