Since the last newsletter we have received a long letter from one of our valued clients who recently upgraded from Unibase 7.2 to Unibase 7.4i. All users who are not frequent upgraders, upgrading from a prior version to a current version of Unibase by DMAC, can probably benefit from my comments on the clients letter. While the comments are specific to Unibase by DMAC, they are probably much more generic and can benefit all people who upgrade.

Manual Not Up-to-date; Use ReadMe File & On-line Help

Unibase by DMAC, release 7.4i, first appeared in the Spring of 1996. At that time, the manual was up to date as best we could make it. By the time a year has passed, significant new features have been added as we go from release 7.40i to release 7.43i.

The readme file is distributed on the installation and upgrade disks. It contains mention of all changes by release number.

For readers unfamiliar with DMAC's release numbering system, read about it in Winter 1994 issue of "The Unibase Technical Review" on DMAC's web site dmac-unibase.com or in the article on DMAC in the September-October 1994 issue of "Work Process Improvement TODAY" magazine.

On-line help with the <info> key is kept current with each release and can be changed by clients to include addition examples and interpretations.

Upgrade Causes Pressure

When an upgrade is begun, everyone is under pressure until items settle out. Many users set up a separate root and convert all of the Unibase environment in that root, make sure it is running before introducing it throughout an organization. This will reduce pressure, but not eliminate it.

No Improvements in Speed

Between releases 7.2 and 7.4 Unibase by DMAC grew to include very sophisticated memory management and a variety of operating systems from a single load and a large percent increase in error checking on the fly and number of verbs and control functions. Running on older (386's in this case) machines is still possible but the improvements shine on newer faster machines.

In the Quality Assurance Process, see article on timing this issue, DMAC checks that the latest release running on a modern computer is as fast as or faster than the previous release but we do not go back to check speed on a prior cpu computer model.

Edit Parsing Was Necessary

Whenever the second digits differ in an upgrade like from 7.2 to 7.4, parsing with the new release presents unique opportunities. The ".r" programs (similar in concept to JAVA runtime code only more stable and older) will probably still run, but when parsed with the new release things happen.

In the past, Unibase by DMAC and its predecessors, tended to be very lax in checking syntax. As more and more features, verbs and control functions are added Parsing becomes more and more strict. What caught this client was the lack of a space (technically white space) separating variables from the divisor sign (/). We can add that on conversion to Unibase by DMAC this space is inserted. We can add a comment, if it is not present, to the manual and readme files about white space between components in a statement.

31-character Names Caused Grief

Of cours, it was the additional capability of 31 character names that made this change become apparent and a parse on the new release did immediately identify the problem.

LANPR1 Does Not Work

Microsoft and Novell are locked in a battle to make each others product difficult to use.  Every time we upgrade a compiler, client, interface libraries or server we have to run exhaustive tests to find which features are being messed up.

The entire spooling concept on Novell Servers has been shot down many times by Microsoft.  Finally in Novell 4.11 spooler functions have completely changed.  We still don't know how we will implement the permanent fix for this change.

We solve these problems when we can; always trying to keep as much compatibility as possible.  If we can find an alternate way to achieve the same result without dependence on Microsoft or Novell -- we recommend it.  This is the reason we are so interested in POSIX and ANSI C and C++ (and other industry wide standards also).

Unibase Needs Conventional RAM

Unibase requires about 500K of conventional ram memory.  On DOS machines loading the network interface packages, and other portions tends to squish Unibase.  We have an old 8088 with 604k of memory upon which we always load the latest release of Unibase by DMAC and make sure it can run.

The process takes dynamic memory overlays and static swapping to run in this environment.  This overlaying is slow.  If 2 megabytes of extended memory are available (requires at least 4 megabytes of memory), Unibase by DMAC will use extended memory.

Keyboard Does Not Work As Hoped

There are so many options for using a keyboard in Unibase by DMAC that I think it is safe to say almost anything can be supported with the various environment settings.  Everybody has a different way of using the keyboard; and each client insists its way is correct.  What is released as standard with using environment variables to customize has changed between major releases.

Thirty One Character Name Overflows

Because the name in Unibase by DMAC can be up to 31 characters long, the change in name length must be accounted for.  We at DMAC feel the longer name space justifies the effort required to upgrade to the longer name space.

Files and Indexes are Bigger

As information for imaging and indexing/sorting 65000 different files was added to Unibase by DMAC, the files and indexes grew by up to double or quadruple on small files.  When distributing data to standalone workstations, users might want to pack data before distributing it.

Summary of Upgrade Snares

By repeating the comments from the user here, I hope we can help others going through upgrades and conversions of Unibase by DMAC.  Please do not take this small list as typical of all conversions or upgrades.

Almost all upgrades in any product today are more difficult than a few years ago.  This is not only about troubles with Unibase by DMAC and application programs; I am speaking about troubles with the operating systems themselves.