[comp.databases] Selecting among Unix Relational DBMSs summary of replies

simmel@ginosko.samsung.com (Sergiu Simmel) (06/22/89)

A while back I posted a request for information for selecting a RDBMS.
Here is the original posting:

+-----------------------------------------------------------------------------+
|I am trying to select a commercial RDBMS product to be used as a back-end    |
|for a system we're working on.  Here are a few criteria for selection:       |
|                                                                             |
|  * RDBMS should have a large enough share of the Unix(tm) RDBMS market;     |
|  * RDBMS should have a quality (stable, well designed, working, complete)   |
|    programmatic (3GL) interface (no pre-processors, such as ESQL, required);|
|  * RDBMS should run on a variety of hardware platforms (Suns, Pyramids,     |
|    perhaps DECstations 3100);                                               |
|  * company should provide good technical support for 3rd party developers;  |
|  * company should provide adequate VAR program and support;                 |
|  * RDBMS should have adequate performance;                                  |
|     * RDBMS should be compact in size and architecture.                     |
|                                                                             |
| Although there is no total order among these criteria, the importance tends |
|   to decrease down the list.                                                |
|                                                                             |
|   Assuming that the result of applying all these restrictions is not null, |
|   what product would YOU select if you were me?                             |
|                                                                             |
| Please direct responses via e-mail to address below.I will post a summary if|
| relevant responses arrive.  Thank you, in anticipation, for your effort and |
| responsiveness.                                                             |
+-----------------------------------------------------------------------------+


Here is a summary of the responses I received.

(1) From: ...!uunet!cs.utexas.edu!sun-barr!rutgers!uwvax!tank!cs_bob@gsbacd.uchicago.edu

    [...] I think your first requirement would point you to
    either INFORMIX, Unify or Ingres, and perhaps Oracle, depending on
    what you mean by "large enough share". I don't know much about Unify,
    but the other three pre-process 3GL programs. Informix has what I consider
    the only true 4GL in the relational world. Both Ingres and Oracle market
    4GL's, but as far as I'm concerned, neither of them are robust enough to
    warrant calling them true languages, let alone 4GL's.

(2) From: c162-aj@zooey.Berkeley.EDU (Mark Christiansen)

        As for some recent light research that I have done on Sybase,
    they are (as of the articles read from 1988) the only database with
    government security clearance, data integrity (ie. non null primary
    keys, and matching foreign keys, etc), and they have an excellent
    data backup.
	The DB may be proted from one VAX on a network to another
    if the original 'goes down' and the DB also uses a mirror disk
    access system where it saves to different places on the disk, or
    different drives simultaneously to protect lose of data to disk
    crashes (or limits the lose at least).
	DataServer (the DBMS for Sybase) also using a multi-threaded
    access to relations so that many users may be using the exact same
    info without integrity conflict (ie. one writer, but many readers)
    and DataServer also can shovel many users into a small area.
    Somewhere on the order of 50 users to 1 Meg (don't have the articles
    in front of me) which leaves massive room for cacheing resulting in less
    I/O.

(3) From: Bill Elgie <uunet!RELAY.CS.NET!elgie@canisius.csnet>

     Sybase would meet your needs re compactness and speed. 
     INGRES comes closest all around, but is definitely not compact.....

(4) From: uunet!lgnp1.ls.com!mdi386!bruce (Bruce A. McIntyre)

    When NCR looked around for a system to use for ALL INTERNAL NON-TRIVIAL
  applications, they chose PROGRESS from Progress Software.  I have tracked
  their programmers and users over a two year period, and they are all generally
  happy with the choice. (Naturally, since NCR is all over the world, I haven't
  talked to more than a representative sample)  We use Progress here because
  we have to make a living doing application development, and can't afford
  short sighted approaches.  Rarely have we found it necessary to use the
  "c" interface, and with the release of 5.5 we will be able to connect to
  multiple databases distributed over a net, of both/either Oracle and/or
  Progress.  Even if you select ORACLE as the database, Progress would be
  an excellent choice for the front end.  While Progress contains most of
  the facilities of a 4GL, it also maintains the control possibilities
  of a 3GL when you need it... Give their test drive a try for only $95.
  It is a FULL DEVELOPMENT SYSTEM, limited only by the number of times you
  can start the server against a database. It is also very transportable
  between all of the environments you mentioned, plus DOS, LAN's, CTOS/BTOS
  XENIX and others comming.. (A/UX and OS/2).

(5) From: ulowell!harvard!SGI.COM!hargrove (Mark Hargrove)

  As a developer a few years ago, I had to make the same decision with about
  the same set of general selection criteria.  At the time, I chose Unify, and
  it turned out to be the right choice.  This was about three years ago.  Today
  I'm responsible for data base technology at Silicon Graphics -- and with 
  a new (and different) set of selection criteria in front of me, I'm making
  a different decision.  BUT -- if I had *your* criteria, I would *still* 
  choose Unify today.

(6) From: uunet!cs.utexas.edu!radian!bloom!bobd (Bob Donaldson)

  Your restrictions may produce the NUL set, but before you give up you
  should check out Empress/32 from Empress Software, Toronto [(416)922-1743]

  It was developed under UNIX and is very well-structured and reasonably 
  compact.  They have a good VAR program (we are in it), although the
  technical support isn't all it could be - we find that we give them
  technical support on occasion.
  The C interface is great - 3 levels
	-   embedded SQL
	-   record/field level navigation
	-   record/field level navigation plus complete control of
		memory allocation and buffering
  The market share could be better, but it seems to be growing.
  It comes with an applications generator that is VERY easy to do easy things
  with, yet VERY POWERFUL.  It takes a while to get used to the powerful
  features so you can use them well, but they include transparent access to
  operating system calls, other executables, etc.  The system also lets you
  bind in your own SQL-callable functions written in C.
  I like it.

(7) <anonymous - by request>

  *#1: Check out Informix, RTI (Ingres), Sybase and Oracle.
  *#2: Check out Sybase and Oracle. Also ask RTI. Ingres v6.1 might or might
       not have a programmatic interface. I don't think that Ingres v5 has 
       it (I don't have the manuals at my desk now).
  *#3: Informix, RTI, Sybase and Oracle.
  *#4: RTI! Oracle is big, but can be hard to work with efficiently. My 
       experiences with Informix and Sybase are quite good, but RTI really 
       stands out.
  *#5: All of the above, but especially RTI and Sybase.
  *#6: RTI!
  *#7: RTI, provided they have a programmatic interface in v6.1. Otherwise
       Sybase.


================================+==========================================
Sergiu Saul Simmel              | VOICE: 508-685-7200x133 FAX: 508-685-4940
SAMSUNG Software America, Inc.  | INTERNET:      simmel@ginosko.samsung.com
 One Corporate Drive            | UUCP:        {decvax!{gsg,cg-atla},uunet,
 Andover, MA 01810              |                   ulowell}!ginosko!simmel
================================+==========================================