Integrate lftp into Solaris [PSARC/2008/276 FastTrack timeout 05/02/2008]

[Garrett D'Amore]

John Plocher wrote:
> Garrett D'Amore wrote:
>> Checking the lftp site, it seems that lftp is perhaps superior to 
>> ncftp in some ways.
>> I wonder though, is it possible to write a wrapper for lftp so that 
>> it can emulate ncftp?   I'm all for taking the latest improvements 
>> from the open source community, but I still think it would be "nice" 
>> to minimize the number of different implementations of what is really 
>> the same 
>
>
> Why?  Or more precisely, what do you mean by "minimize the number of
> different implementations"?

I meant "that are integrated into Solaris".

And the rationale is quite simple -- more code in Solaris == more effort 
to sustain, support, larger install image, bigger download, generally 
more "bloat".

If having multiple different implementations has some core value (e.g. 
OpenSSL and PKCS#11 both have many feature overlaps, but we really do 
need both to support the different kinds of applications) that's fine.  
But if users or developers don't gain a noticeable benefit by having 
multiple implementations, then we (Sun/Solaris/OpenSolaris) gain a 
benefit by having fewer.

>
> We can't go off and disband a bunch of FOSS projects that we don't like,
> and we can't stop people from getting/compiling/using them...

Again, I'm just talking about what we deliver.  Sorry if that wasn't clear.

>
> All we can do is to  make it harder for users to actually get binaries
> that run on OpenSolaris for these things.  And every time we do /that/,
> we make it harder for users to take OpenSolaris seriously because it
> does not meet *their* needs.

I'm asking if we can meet the needs of these users with a single 
implementation.  Not trying to prevent anyone from using their favorite 
tool or syntax.  In other words, this is more like having kmdb emulate 
older mdb... if lftp (or ncftp) can serve as a stand-in for one or more 
other file transfer implementations, then I'm all for reducing the total 
number of implementations we ship.  But they can't fully or at least 
adequately stand in (via emulation, shell wrapper, or somesuch), then I 
withdraw this point altogether.

>
> In my mind, EVERY SINGLE FOSS project out there[1] should be available
> in a repository, in ready-to-run binary form, waiting for someone who
> needs it to download on their OpenSolaris system.

The Repo argument is fine.  But that's not what this case is about.  Its 
about what is *integrated* into Solaris.

>
> Now, if you were saying "only *some of them* should be considered
> candidates to become a core part of the OS, we might have a different
> conversation, but even then, if we had a full mondo-repository, why
> would it matter?

If we had a full mondo repo, then no, I don't care either.  At that 
point, more is probably better just for sake of variety.  (Assuming that 
we don't have to fully and equally support everything in the 
repository.  Support isn't free.)

    -- Garrett

>
> -John
>
> ____
> [1] that makes sense for Solaris - there are obviously quite a
> few that don't, such as MS-Windows editors, linux-kernel specific
> admin tools...  :-)