On Wed, 1 May 2002 15:45:19 -0700, Stan Sieler <[log in to unmask]> wrote:
>Re:
>> > Even *HP* admits the problem: they offer migration consulting to
migrate from
>> > 10.x to 11i.
>>
>> Huh? Just what is it that HP is changing so much that forces significant
>> application changes? Unix isn't Unix? I'm no HP-UX expert so I don't
know
>> if this is simply poor planning or if there is some logic behind the OS
>> changes that require application changes!
>
>Things like:
>
>1) dropping software:
>
>Some examples include:
>almanac getcollate
>getlangid getlangs
>gettables nlappend
>nlcollate nlcollatesys
>nlconvclock nlconvcustdate
>nlconvnum sub_in_outstr
>extract_nldt format_nldt
>get_end getjulian
>isvalidgregorian isvalidjulian
>nlcrtm calendar
>clock nlfindstr
>nlfmtcalendar nlfmtclock
>nlfmtcustdate nlfmtdate
>nlfmtlongcal format_outstr
>nlfmt_cleanup
>
>
>2) moving software:
>
> One at random from ITRC:
> … Some libc APIs were moved to libm, libnsl due to standards/de facto
> standards and will require a link line change to find the APIs in the
> new library. (This breaks some Makefiles)
>
>3) command changes:
>
> One at random from ITRC:
> "What is the HP-UX 11.0 command is equivalent to the
> HP-UX 10.x
> command cmgetconfig-f ?"
>
> "find" command parameters removed
>
>
>The release notes for 11.00, at
>http://devresource.hp.com/STK/partner/relnotes/11.00/relnotes_1100.txt
>provide a good starting point for answering the question :)
I think that Stan has made my point about UNIX incompatibilities quite
clear. UNIX has never respected and does not value compatibility and UNIX
users (of all people) should know this. (I do not know for sure why this
is the case, but accept the fact that it is. I suspect that it is a legacy
from the university laboratory days of UNIX which was never intended to be
an operating system for business use.) Potential new UNIX users coming
from MPE need to understand that this is a major philosophical change
between MPE and any UNIX; i.e., there is no reasonable guarantee that
applications will not be affected. HP tells you this also.
In addition the myth of UNIX "standards" is a joke. UNIX from HP, SUN and
IBM (all proprietary) have many similarities, but they are not compatible
with each other! I just returned from the annual user conference for our
application package and the new releases of the vendors software have very
different install experiences from platform to platform (SUN being the most
difficult).
Why do package vendors supply different editions of the same release of
their software for the same platform if the OS is transparent? We found
this to be true just going to 10.20 HP-UX from 10.01, and it has been true
for every release I have ever seen or investigated.
I am not saying that ALL applications will ALWAYS be caught up by a new OS
release -- absolutes are rarely true. In general, packaged software will
probably have a higher chance of this than custom software -- at least
that is my experience with packaged software. But I will make this
absolute statement: ALL applications need to be considered and tested
thouroughly when making a UNIX OS change. To be prudent this should be
done under MPE too, but problems are much less likely to exist for non-
privileged code. To just upgrade the UNIX OS without considering this is
irresponsible.
The HP document that Stan has kindly referenced is very daunting and should
scare the hell out of anyone. Just the table of contents is exhausting. I
have no idea how many pages this document is, but I can say that I have
never had to wade through something even remotely as complex for MPE
upgrades. This is one reason I have an HP S.E. to bring such issues to my
attention and help me steer through the mine fields. Greg says he has
never heard of any 10 to 11 OS issues, but that does not mean that they do
not exist -- it just means that he is lucky so far.
By the way, I have been fascinated to see the discourse over my statement
about file locking. I was referring to the fact that the *operating
system* does not offer this service. Obviously applications or databases,
such as Oracle, manage to accomplish this through language libraries or
whatever. UNIX, as is often the case, pushs functions performed by MPE
down to the application level to handle. The need to internally structure
files with application software is another good example.
Somehow this point has gotten mangled into discussions of being able to
blast away program files as though locking for data integrity had anything
to do with it. These are two completely separate concepts that have
absolutely nothing to do with each other. I was not even remotely thinking
about the idea of replacing programs, etc. with the "rm" command. I have
been using that for years on MPE for UDC's and programs. PURGELINK is
really quite dangerous and I have created a special UDC to control it.
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
