[Info-vax] wrong file format

[Stephen Hoffman]

On 2021-01-07 16:49:23 +0000, Arne Vajhj said:

> On 1/6/2021 9:14 PM, Stephen Hoffman wrote:
>> On 2021-01-06 23:33:59 +0000, 1tim.lovern at gmail.com said:
>>> that is because the philosophy of Unix, at least in the early days, was 
>>> to treat everything as a stream. This made the I/O the same for files, 
>>> devices, whatever. At the time it was pretty elegant. You had a unified 
>>> view of the world.  VMS and other operating systems took a different 
>>> approach and differentiated between files and devices. They also wanted 
>>> a presence in the commercial space, so the idea of using ISAM as the 
>>> native file system made a lot of sense. Having the OS know about file 
>>> structures made applications easier to play together.
>>> 
>>> Unix makes every application have to know about the metadata for a 
>>> file, VMS manages the metadata for you.  You can discover all kinds of 
>>> things about a file simply by asking VMS to open the file and fill in 
>>> all the data structures, FAB, XAB, etc.  Unix will tell you if you 
>>> opened the file and if you hit the end of the file.
>> 
>> What's called Unix is fairly varied, and most of the distros are past 
>> the point when apps need to consider record-level activities or 
>> file-level metadata.
>> 
>> Past the universal debugger printf tool, that is...
>> 
>> The file system and metadata underpinnings analogous to RMS are widely 
>> available, or are integrated.
>> 
>> I find the local Unix systems much easier to work with than OpenVMS, 
>> with lower coding effort, far better development tools, and better file 
>> system support.
> 
> Almost everything is available on Linux. They have the benefits of 
> being big dog.

My Linux usage is largely involving Kali, and I'm not doing much 
development with Kali.

When not working with OpenVMS, mostly with Unix distros, as stated.

Not nearly as often with Linux distros.

>> Reading and writing app data into those files is vastly easier, whether 
>> for loading and saving "preferences"-style data, marshalling or 
>> unmarshalling app data, or using the integrated SQL support.
> 
> Lots of relational databases are available for Linux.

Linux has databases? Staggering. Amazing. Wow.  🙄

> There are not as many available for VMS, but there are several 
> available (Rdb, MySQL/MariaDB, Mimer, SQLite, HSQLDB, H2, Derby etc.).
> 
> If someone on VMS want to use a relational database, then it is possible.

ibid 🙄

>> I haven't had to mess around with file system data structures or such, 
>> and—thankfully—haven't had to deal with anything as involved as RMS and 
>> its interfaces. And with better results.
>> 
>> With OpenVMS, I'm finding it easier and faster with various apps to 
>> bypass RMS and access the storage as sectors—stream access might be 
>> easier, but sector access are what we have available.
> 
> ????
> 
> SYS$QIO(W) with IO$_READLBLK or what?

$crmpsc, usually.  Other options include $io_perform, depending on the 
local requirements.

>> OpenVMS not having an integrated relational database, as is the case 
>> with the local Unix distro...  Ugh.
> 
> I don't think Linux got an integrated relational database.

Some Linux distros do, some don't. Same for the various Unix distros. 
The local Unix distro does include a relational database, and related 
frameworks built on it.

> Most distro's probably offer a few of them (MySQL, MariaDB, PostgreSQL, 
> SQLite etc.) in their package repo.

Approximately all, but yes.

> But as mentioned above then relational databases are also available for VMS.

So?

The local Unix systems remain easier to work with than is OpenVMS, with 
lower coding effort, far better development tools, and better file 
system support.

> Would it make a big change if VSI included a bunch of them as options 
> as part of VMS "distro"?
> 
> I really doubt it.

Immediately? No. Incorporation across even a few existing RMS 
indexed-file apps will take years or decades. Where this and other 
enhancements gets interesting is with overhauls and new work. And with 
what all of the we-love-how-RMS-allows-other-apps-at-the-data-store 
folks now have a standard database they can use and access, and without 
having to try to make RMS into something it's not good at.

For some details on what it's not good at, see the PCSI database, the 
cluster authorization database, etc.

Some resistance to updates and enhancements is to be expected, of 
course—alas, that usually means wads of RMS files.

Put differently, databases are not a short-term investment, nor one 
quickly adopted. And some sites will never adopt the added database(s). 
But incrementally, some folks will adopt.

Because there's a great big gap between RMS and Oracle Rdb.

> Anyone on VMS that want a relational database should be able to get it 
> quite easily.

And the local Unix system will—much like the vastly better IP 
integration over on Unix—remain easier to deal with.


-- 
Pure Personal Opinion | HoffmanLabs LLC