[Oberon] Train VS all terrain truck

[Lars]

On Tue, February 9, 2016 8:15 am, Jörg wrote:
>
> Coming to your niche question: In my point of view, let Oberon run on
> RISC5,
> ARM or Intel. Whatever might fit your taste.
> The benefit of Oberon (not only the language but also the whole Oberon
> system) is that it is so small. So, programming an embedded device (=small
>  memory space and very low power, not too fast CPU) in Oberon could be
> beneficial. You don't have to develop in assembly language but can use a
> rather high level, strictly typed OOP-ready language.

I'm not so sure that OOP is the way forward.

I'm thinking actually that Relational oberon would be the way forward, as
everyone these days reinvents the database by using plain files, or by
using TStringList(delphi), or by using linked lists, arrays, maps (golang,
python?)

It seems to me that a lot of the work done on these computers is dealing
with data, and how to access this data. Pascal/R failed, because it was
ahead of its time back in the day and no one  understood the significance.

XML is a joke, it's worse than old microsoft INI files or unix CONF
(config) files.

Where's the relational languages? OOP tries to make you store your data in
objects, but then you have to map those objects to a real database
eventually, and that becomes an ORM (object relational mapping) mess.

Wirth has never done any work with the relational model, but Dijkstra dug
into it a bit.  It's too bad more discussion on the relational model
didn't exist in Dijksra's papers, and more research. Fabian Pascal has
taken up the job of discussing the relational model, along with Date and
Darwen. However, they are fading into academic obscurity, while projects
like Microsoft Linq make headway. But MS Linq is not a product that a
language can implement, it's a microsoft only tech for C# and .net.  T

o implement something like MS Linq in oberon, delphi, or golang, would
require inventing something similar to MS Linq, but not the same.
TutorialD and DuroDBMS are fading away into academic obscurity just as
Pascal/R faded away into hard drive security. I can't even find
information on Pascal/R it's so dead and unused. One time I found some
information on it.

Also, implementing the relational model into a language means a more
complicated compiler.  However it's kind of like implementing a procedure
or function: at some point you just have to implement a feature that is
useful, you can't just omit procedures or functions and make people
program in one long simple code block.  The problem I see with TutorialD
is it's type system wants to allow everything under the sun, all kinds of
types and types and types will be able to exist. Once again if only
Dijkstra was around as he was interested in Type Theory and such topics.

I don't think OOP is the best way forward, and I think Wirth has skipped
the relational model research entirely. I once emailed him about it and
got no reply, long ago... he has no blog, no way of communicating, it's
all one way communication: he writes books, we are to read them without
any questions. The only way the man can be talked to is to meet him at one
of those programmer wine and cheese conferences where he is handed a
reward.