[Samba] Samba4 join xp problems

[Nico Kadel-Garcia]

On Sun, Mar 8, 2020 at 4:36 PM Piviul via samba <samba at lists.samba.org> wrote:
>
> Il 07/03/20 21:27, Reindl Harald ha scritto:
> > but given that such machines must not be in the normal environment
> > anyways why do you need the latest and greatest samba?
> but the same user can work on an old winxp or latest win10 machine and
> have to share and access the same files that have to be shared between
> the network. So old and new have to coexists

Having to maintain old environments, or upgrading them to something
contemporary, is an old, old problem. It's the worst when the old
hardware can no longer be obtained, and the backup media can no longer
be read. Been there, done that, with ATA drives no longer manufactured
and magtapes from NASA data no longer legible on contemporary, higher
mag-tape drives, and fiscal data stored on old SCO OpenServer systems
for which intervening license and software updates were no longer
available. The data involved was mortgages and other real estate
documents, involving 30-year mortgages, so it *had* to be accessible
for decades.

Samba access to fileservers was critical. Maintaining compatibility
with newer file services and access control was not trivial, and it
was not cheap in consulting time, and I was permitted to publish my
notes. So I absolutely sympathize with your difficulty.

> > use RHEL8 or something with 10 years support and plug the network with
> > the xp machines on a second NIC for security
> >
> > and 15 years: forget it, at least if you care for security update son
> > teh server
> yes but it is not so simple. Some service has to be upgraded for a lot
> of reasons but can't because breaks some compatibility. When a retro
> compatibility is broken there are a lot of aspect to evaluate...

Oh, dear lord, yes. This used to be a big chunk of contracting work
for me as I was doing international travel to support my ex-wife's
career.

> > i still need to understand why the most expensive stuff is the sloppiest
> > crap and that "high value machines for tens of thousands dollars" can't
> > be controlled by anything not older than 20 years
> because old 20 years machine can do the same work than can do new
> machines. Yes frequently happens that new machines can do his task
> faster or can do some task more in the same time and that seems to be a
> good reason to upgrade them; but if you don't need this more fastness or
> the more tasks they do, you have to upgrade only because the OS of the
> PC that control it is too old... or often happens that a new machines
> are placed side by side with the old ones and the olds one are employed
> rarely but when them are employed they have to access the same network
> and access the same shared data...

The same is very true in academic environments, where money for
upgrades is very limited. There is also a real reluctance to introduce
the leading edge of the "bathtub curve" for failures of new system.

> Any way I can understand that for developers retro-compatibility is a
> hard task...

It's one of the things that DevOps personnel get paid for. I am
eternally grateful to the older engineers who *tolerated* and educated
me in existing technologies as a new sys-admin back in the day, and
hope I paid off their investment with introducing them to tools like
Samba for multi-platform work back in the 1990's.