Re: New RDBMS implementation

On Tue, 04 May 2004 10:13:17 -0300,
=?iso-8859-1?q?Leandro_Guimar=E3es_Faria_Corsetti_Dutra?= <leandro_at_dutra.fastmail.fm> wrote:

>Em Tue, 04 May 2004 01:54:40 +0000, Alfredo Novoa escreveu:
>
>> And it is probable that they can not implement the Transrelational Model.
>
> After talking this with Nathan, I am not so sure... the issue
>seems to me more that he doesn't quite believe the thing. Nor do I
>can say to have quite understood it... someday I need to read those
>patents.

I don't like the part of insertion and deletion. I am using a very different approach based in cache conscious B+Trees or B*Trees

>>>OTOH it is a complete
>>>RAD environment.
>>
>> The frontend is not flexible enough for us
>
> Interesting, what do you need above and beyond what Dataphor
>provides or intends to provide?

A more traditional RAD approach. I need total control on the GUI, including custom visual controls.

>> And we need [...] cheap server licenses
>
> Nothing could be cheaper than using PostgreSQL as a device,
>granted they will probably charge once they have their own thing.

Does that device exist?

But it is not very comfortable to use PostgreSQL with windows.

>> and preferabily a distributed DBMS
>> architecture with many data cached in the desktop PCs, compressed and
>> encrypted communications, etc.
>
> They do some client constraint enforcement... data cached on
>the client is an interesting concept, it is also a tall order.

It is not very complex if you have a distributed DBMS server in the client, or something equivalent.

>>> Did you actually read the patents?
>>
>> Of course
>
> This 'of course' just proves how higher are your powers of
>concentration and comprehension than mine!

I said of course because it would be stupid to try to improve the patent without reading it :)

Some parts are really boring. You can skip many boring algorithm descriptions if you only want to get the fundamental ideas.

Do you have the .pdf version? It is a lot more readable and it has many drawings.

>> but my data structures are rather different and they can be
>> used in disk based DBMSs. Date mentioned that the Transrelational Model
>> was adapted to be used with disk based DBMSs and I wonder if they used the
>> same solution as me (modified B-Trees).
>
> So you're in touch with The Man... good.

No, I meant I don't know that. I never had an exchange with Date.

>> I also wonder if the modifications are enough to avoid problems with the
>> patent if I want to distribute it in the US, Canada or Japan.
>>
>> Software patents are crazy.
>
> Agreed... had you the money, it would be interesting to get
>lawyers to clear up the path to you, or to arrange for a license with
>ReqTech -- but according to Nathan they're not quite responsive.

I don't live in one of those three countries. It is not an important problem by now.

> There are two concepts to 'portable' here... what is
>potentially portable, and what can run elsewhere now... Dataphor can't
>yet.

Yes, you don't know if it is portable until you try to run it.

>> The published LARL grammar has many mistakes, it is a toy language and
>> some constructions are ugly IMO.
>
> I seem to remembered this was admitted by the authors
>themselves...

The mistakes and the ugly constructions too?

I sent a version with many fixes to the author (a swedish guy who works for Mimer), but I think he misunderstood it.

>> start transaction and commit transaction are not required. Any language
>> script is executed in the same transaction by default.
>
> So you implement that idea of the statement as the
>transaction... I haven't yet made my mind on it, but as long as one
>doesn't deal with a host language and a data sublanguage it really
>seems the thing to do.

A statement list as the transaction. The only problem is when the list is too big to fit in memory.

> Now what if one needs a data sublanguage in a host language?

You can build a complex script and to send it to the DBMS.

>> The language is case insensitive and it does not have '_' characters
>> inside the keywords. Like in SameTypeAs();
>
> Do you mean you forbid that, or just that you've gone
>Hungarian yourself? Bad or not so bad...

It is not forbiden but I don't use '_' in the builtin operators. Of course it is a valid ident character.

>> It has more builtin types like: numeric, date, timestamp, int32, int64,
>> int8, etc.
>
> As long as you enable user defined types, possreps and the
>whole shebang...

Int32 is a subtype of int16 and int8 is a subtype of int 16.

Integer is a subtype of numeric and so on.

>> BTW what do you think about Tutorial D as a general purpose programming
>> language?
>
> It was never meant to be one.

But it is a lot better than the current languages in many ways.

> The authors themselves have
>stated that, and that it would be interesting to see how would be a
>functional D...

I am very skeptical about building business information systems without variables.

>> The most important problem is the IDE :(
>
> Yes, the big issue for me to experiment with Dataphor is that
>VisualStudio.Net is required.

I meant that is very difficult to implement a good IDE or to integrate a new language in Visual Studio.

> Java would get you the portability .Net is yet only
>promising...

Almost everybody I know is moving to .Net.

> I'm not a coder, but I gather Java is not significantly
>worse than .Net.

I disagree.

> If you really meant to use an existing SQL DBMS, PostgreSQL
>would be the way to go, perhaps. It was once relational (Ingres
>QUEL), and it has already been morphed from QUEL to SQL, so it should
>have yet the interfaces clean enough to go back.

IMO it is a lot better to start from scratch. I could spent many months trying to understand the code and I would be anchored to an obsolete architecture forever.

>> A bit soon for that, but my associates would kill me if I publish all the
>> code :)
>
> Perhaps they should learn about copy rights... publishing does
>not mean abdicating of any rights.
>
> There are quite some companies making big bucks by publishing
>code and then supporting it. Zope, the GNU toolchain, the GNU Ada
>compilers, Ghostscript all work like that.

MySQL could be a better example. I have to think about it, but I was always skeptical about open source.

For instance Borland published the source code of Interbase and now they are losing a lot of sales due to Firebird.

> I hope to include Dataphor in a technology evaluation for my
>employers later this year. Perhaps you will have something
>publishable by then?

I hope that something usable, but the priority is to have a DBMS for ourselves. Manuals and a nice Web page is not a priority.

Regards
  Alfredo Received on Tue May 04 2004 - 17:04:32 CEST