pg.ddx.ioUniversity INGRES on Usenet (1985-1995): threads about Berkeley INGRES, QUEL and EQUEL from comp.databases.ingres, comp.databases and net.database
From: jamesl@netcom.com
Subject: Re: ODBC competitors?
Date: 1995/08/23
Message-ID: <jameslDDqxrC.4J1@netcom.com> (raw)
References: <0kCAeP200jcQ02zls0@andrew.cmu.edu>
<41cn8s$i8k@knot.queensu.ca>
KeJin wrote:
: Some facts and *my* opinions about ODBC:
: (1) the CLI specs in draft version of ANSI-ISO SQL standard(would be
: SQL 96) is more or less an ODBC or ODBC is more or less an ANSI-ISO CLI
: (especially for ODBC 3.0). The ANSI-ISO CLI is language independent
: which can be implemented and called by most(if not all) languages
: James Logajan has mentioned. Alos, unlike on of the poster has said,
: ODBC is NOT C oriented. For example, the concept of 'not null terminated
: sting' in ODBC is introdced for languages other than C such as Cobol, Basic,
: etc. I believe most C programmers(at least me) are extremaly 'hate' this
: concept. ODBC is not a protocol. The reason ODBC is specified not as
: a protocol is, as a standard, it need to allow application to do database
: access with and without network involvement. However, driver(2-tier and up)
: implementers can use any protocol they like or available.
Note: my comments, to which Ke Jin is replying, were directed entirely at
ODBC 2.0. I do not have any information on ODBC 3.0. If you would so kind as
to tell me when the ODBC 3.0 spec came out and relevant purchasing information,
it would be greatly appreciated. The reference document for ODBC 2.0 (at least
the only one that I'm aware of) is:
"Microsoft ODBC 2.0 Programmer's Reference and SDK Guide: for Microsoft
Windows and Windows NT" 1994, Microsoft Corporation, ISBN 1-55615-658-8,
$US24.95, UK21.95, Canada$32.95.
Here is the very first line of the very first section "About This Manual":
"The Microsoft(R) Open Database Connectivity(TM) (ODBC) interface is a C
programming language interface for database connectivity."
No mention of Cobol, Basic, etc. occur in the index. Note the trademark
on the phrase. An oxymoron worth savoring: "Open Database Connectivity(TM)"
Also, the issue of 'not null terminated string' and the way Microsoft chose
to pass strings to their SQLxxx() functions has absolutely nothing to do
with any attempt on their part to be language neutral. If you study the
history of Windows and examine the Windows system calls that require
strings, you find that this is a hold-over from Pascal calling conventions
for strings. (If memory serves me correctly, Microsoft Pascal places
arguments on the stack in reverse order from C. And since Pascal strings
use a length/string convention, rather than the C string/null byte convention,
I think you'll find that layout of Windows SQLxxx() arguments suddenly
makes sense. Microsoft was only making life easy for them, not others!)
: (2) ODBC actually include two parts. i.e ODBC CLI and ODBC SQL. the ODBC
: SQL of cause is very SQL specific. However, ODBC CLI is not. There
: are nonSQL driver available (such as ISAM driver) which allows
: user to access flat file through ODBC CLI.
I think you contradict yourself here. I could be wrong, but just how would
a shrink-wrap application know the correct syntax a-priori to query a
non-SQL driver? One of the points of ODBC is to disengage the app
from the specific engine via a standard query language. I take exception
to the query language chosen (too narrow), but do understand the need for
a standard query language. If you could give me an example of what you
you mean, it would help.
: (3) Sun and HP will soon bundle an ODBC driver manager with the operation
: system(Solaris and HP/UX). That is, when you get the os(i think most
: reader know they are unix), you will automaticlly get an ODBC driver
: manager(as the standard system library). I believe other unix and
: workstation OEM will follow. Many database vendors and third party
: companies offer ODBC drivers or Gateway on unix, vms, mainfram, etc
: Much samiliar to ANSI CLI which can be said is 100% platform independent
: specs, 99% of ODBC is platform independent. gnu iODBC driver manager
: (which is a ODBC 2.0 compliant ODBC driver manager) is mainly targeted
: for unix platforms and already ported to SunOS, Solaris, HP/UX, UnixWare,
: Irix, NCR, Linux ELF, DEC Alpha, etc.
The iODBC that you posted (and authored, according to the documentation)
is impressive, although I haven't had a chance to compile and test it.
Unfortunately here is another problem you have: a couple of your header files
appear to be edited copies of the Microsoft headers sql.h and sqlext.h -
these are both copyright by Microsoft. I do not know if you acquired
permission to copy them, but you should have at least edited the comments
and/or re-arranged things more if you did not. I seriously doubt that
you can put them in the public domain via GPL if you or Empress did not
acquire the right to do so.
: (4) Some database vendors(include Infomix, IBM, Watercom, etc.) have
: adopted ODBC as the native CLI for their database engine(s). The reason
: is not because of ODBC or at least not because of it only. It is the
: first prepared step for adopting ANSI SQL 96 CLI specs. No matter where
: microsoft may jumped to, ANSI-ISO specs will be accepted by most
: (if not all :) ) database vendors. Thus, if ODBC(esspecially ODBC 3.0)
: is a established industral standard and covers(at least
: most part of) ANSI-ISO specs, why don't use it before the publication
: (and adopt) of the ANSI-ISO specs.
Answer to your last question: because Microsoft owns the trademark to ODBC
and the copyrights to all the header files needed to compile apps, drivers,
and of course the specification of ODBC, it can be expected to defend its
operating system from competitors by denying them the use of its standard
at the most damaging time. (Like when they have been suckered into adopting
it and are vulnerable).
: (5) http and html is very close to something described by James --
: not language(i mean programming language) specific and be a protocol,
: not SQL specific, not tied to mw-windows or other os, not coming from
: microsoft -- the http and html. However, when a http server need to
: access a local(i.e. on the same mechine) database, it is much efficient
: to do a direct CLI access rather than through another protocol
: gateway.
: Ke