[Info-vax] Still no DIR/SORT_BY_TIME

[Stephen Hoffman]

On 2015-08-15 14:56:40 +0000, AEF said:

> On Wednesday, August 12, 2015 at 6:55:22 PM UTC-4, Stephen Hoffman wrote:
>> On 2015-08-12 22:29:26 +0000, xxxxx said:
>> 
>>> ... I'm just surprised that after ~38 years someone at DEC/... would 
>>> have gotten around to it.
>> 
>> It's certainly reasonable -- maybe as part of that newer file system.
> 
> Newer file system?

See previous VSI discussions — they're planning on providing a path for 
folks on ODS-2 and ODS-5 that might want to migrate to something 
slightly less limited, less encrufted and better performing.

>> Or you can put the timestamp in the filename.
> 
> Too much work.
> We are running a file-transfer system with files coming and going on a 
> daily basis. We add our own time stamps. In some cases the customer is 
> nice enough to put in _their own_ time stamps. And we have to add our 
> own anyway!

Are files even the appropriate container for that?

>> But then I'm increasingly working a whole lot less with writing or 
>> storing zillions of files -- and trying to create or store even fewer 
>> -- and a whole lot more with data stored in databases.
> 
> I'm migrating a file-transfer system. And there's a real crap-load of files!

If it's typical path, what you're porting was a simple design that was 
well suited to its intended environment, and has since been scaled way 
past sanity and efficacy, and it's undoubtedly become entwined 
throughout.


>>> And then write a nice front end for the database.
>> 
>> It seems maybe this is more about the Oracle database than the 
>> operating system platform?
> 
> Most of the work is setting up jobs on the Unix/Sybase side. And most 
> of that is entering data into the database that tells the 
> well-established perl scripts what to do. Each and every file must be 
> carefully processed and tracked, with numerous things happening 
> differently for different jobs. Some are pgp-encrypted. Protocols vary 
> (ftp, sftp, NDM, etc.).
> 
> And AutoSys runs the show; there's that too. Once the job is set up on 
> the Unix side, we simply stop the corresponding job on the VMS side. 
> Then something breaks. We go back to the VMS side, fix the Unix side, 
> and then do take 2.


So you're also rolling your own a distributed scheduler, too?


-- 
Pure Personal Opinion | HoffmanLabs LLC