Last Friday, I had the privilege of traveling to Renton (Seattle),
Washington to meet with Wayne Erickson and Razzak.  Along with my
programming associates, we had been using R:Base 6.5++ for Windows build
1.840K on large complex databases with sophisticated R:Base and Tango
applications.  In general, 1.840k is a superb product.  It is chalk full of
new features and significant bug fixes.  Because of the complexity of our
database environment - tables with a primary key and several unique keys,
all referenced; tables with multiple compound foreign key references (up to
four columns) against unique keys, tables with a single columns
participating in multiple compound foreign keys each referencing different
tables; extensive use of varchar and varbit columns, lots of multiple user
pressure from concurrent native R:base Windows and DOS applications and
Tango applications, in some cases from world wide users - we were
encountering a few obscure and one nasty problem.

Specifically, when adding a column to a multi column table which had a
varchar column as the last column in the list and the added column was
appended on below the varchar column, R:Base would error when you tried to
assign a constraint such as a foreign key or a primary key to that column.
Second, RBdefine could corrupt the sys_indexes table when a column with a
primary  or foreign key constraint was dropped.  Using the alter table
command at the R> operated correctly.  Third, under certain circumstances,
using RBdefine to delete a foreign key reference for a compound key in which
one or more of the participating columns participated in other foreign key
references would also error out.  Again, using the alter table command at
the R> did not produce the error.  Finally, we were observing the
degradation of autonumber integer columns under heavy Tango use, where such
use resulted in lots of inserts.  The autonum degraded in a way such that
its value in the sys_defaults table became null.  In Rbdefine, displaying
the column with the autonumber yielded a display of a real number xxx.yyy
when the datatype was an integer.

I have been an R:Base programmer for a long time - maybe 18 years or more.
I have always been a prolific beta tester.  I have experienced the hay-days
of Microrim, the low days of Abacus, and the Glorious days of RBTI.  I have
run all the good versions of R:base (which ironically all tend to be even
numbers (2X, 4X, 6X) and the less than good versions (3X, 5x).  I always
found Microrim to be reasonably cooperative; I have always found Razzak to
be extremely cooperative.  This weekend, however, was amazing.

Wayne and I simply sat down, in a hotel room, he with his laptop and I with
mine.  I showed him a problem, he reproduced it on a copy of one of my
databases housed on his computer, and he then tweaked the R:Base source code
to eliminate the problem, issued a new build of R:base, put it on a zip disk
so it could go on Razzak's computer, from which it went on my computer.  By
shortly before midnight, we had 3 out of the 4 problems put to bed.  Wayne
got up at the crack of dawn the next day and while Razzak and I were eating
breakfast, he fixed the last problem and we had the final build waiting as
soon as we got back to the hotel room.

I am relating this experience in some detail, because, I believe that it has
some truly profound significance.  First, the reason Wayne only fixed four
bugs is that that is all I could show him.  I don't know of any more.  I
suspect that if I had 5 or 6 or 10 or 12 replicate-able bugs, he would have
fixed every single one of them before noon the next day.  I would like to
encourage everyone to make use of RDCC and document anything that is wrong
with R:Base.  Be very careful to document it in such a way that it can be
reproduced.  If you can send a snippet of code or a portion of a database,
do so.  It is my experience, that WITHOUT EXCEPTION any reproducible bug you
document will be fixed.  The intimacy of this kind of relationship between
the development community and the database owner and code author is
unparalleled.  It is clearly one of R:Base's great strengths.  I would
challenge any other developer of any other database to relate an experience
wherein the database owner and code author are so approachable and so
helpful.

Second, and this should not be underestimated:  it didn't take Wayne very
long to fix any of my bugs, some of which were tricky.  I got the distinct
impression that Wayne knows R:base source code like the back of his hand.
Wayne is simply an irreplaceable and invaluable asset to the R:base
community.  It is apparent that he "knows his onions" and likes to "hand
around in the garden."  Moreover, this is not the first or the second or
even the third time that Wayne has helped me out before; it is the first
time that I have had the pleasure of working one on one with him.  I have
nothing but admiration for his ability and his willingness to help.

After Wayne retired for the night, I FTP'd build 1.841 to one of my large
client sites and upgraded them on the spot (Using VNC and the hotel room
high speed internet connection).  I even executed a remote reload of the
500+ Meg database, all from my hotel room - a tribute to the manageability
of the R:Base environment.

Razzak then gave me a tour of all of the new features in build 1.841 (go to
the R> and type show version to see what build you are currently running).
Some of the features were quite cool.  I think the official version of
R:base will ship around June 15.  I plan to upgrade all of my clients
immediately - this will easily be the strongest and best Rbase that has ever
shipped.

I then tested my luck and asked Razzak to show me R:base 7.0.  He did.  From
what I can see, it will be a dramatic improvement over 6.0 in both look and
feel and functionality.

As if that wasn't enough, Wayne also tweaked Oterro and gave me a new build
of it as well.

I think the future of R:base is very very bright, due to Razzak, Wayne, and
the special relationship that they have extended to the R:base development
community.  R:base 1.841 is a fabulous product.  Thanks Wayne, Thanks
Razzak.

Harlan







Reply via email to