Posters to the HP 3000 newsgroup/mailing list on the Internet got an early start on several topics in the month of January, when a record number of messages crossed between members of the 3000 community. (Our count showed more than 1,550 for the month). Community members found time to talk up taxes, Year 2000 problems already happening with credit card systems, travelling with laptops, and once more got wistful about old computers they loved but are now long gone.

The high volume of personal traffic didn’t elbow out some serious technical discussions, though. And readers got a mix of the two when more than a dozen programmers testified about their favorite HP 3000 editors. Programs which got votes and are built for or run on HP 3000s included HP’s EDITOR, QEdit and Qedit for Windows, Whisper Programmer’s Studio, PSG’s Facade CoEditor, QUAD, HPEDIT, SpeedEdit, RPG Interactive Screen Editor (an HP product that interprets the Card Type columns). Several wizard-level programmers even admitted to regular use of vi, an editor responsible for both magic and anguish that’s been in MPE/iX for several releases by now. Coupled with a discussion of where to find “good business software for the 3000,” the two threads showed the nature of the list: a place to share what you believe as well as what you’ve learned.

Determining datasets’ MPE file names

Stephen Pennine was replicating an existing dataset within a database and adding fields to a corresponding set in another database. He needed to know the dataset name to assess disk space requirements, and help arrived in both long and short answers. Mike Hornsby of Beechglen Consulting answered at length, and one trick was to “redirect a QUERY FORM SETS or a DBUTIL SHOW DBNAME CAPACITY into a file and trim away all but the set list in EDITOR. A LIST ALL will then have the line numbers corresponding to the set numbers. (One must be careful to use the ; password in QUERY because you can’t see datasets that you don’t have access to!)”

Hornsby added his own favorite method, “to do a home clear interactively just after typing show dbname capacity in DBUTIL, then I hit return. You have to delete the headers, then the result is that dataset number is the screen line number when you put the cursor on the line that has the desired dataset name.”

Stan Sieler said that “determining name of a database root file, given the open database “handle”: DBINFO mode 406 returns the file number of the root file, and the root file name. Given the name of the root file, and a dataset number, generating the dataset file name [can be done with ] various methods, ranging from calling HP internal routines to taking the root name in HFS format (e.g., /MAIN/PUB/SALES) for sets 1 to 99; appending two-digit setnum (e.g., set 7 is /MAIN/PUB/SALES07) for sets 100 to 109, appending “A#”, where “#” is set# - 100 (e.g., set 107 is /MAIN/PUB/SALESA7) for sets 110 to 119, append “B#”, and so on. This isn’t complete, as it omits the jumbo dataset discussion and the b-tree discussion.” Thomas Cook added that Adager or DBGeneral can reveal the names through a schema decompile.

As always, an MPEX user showed off his shortcut. Ted Ashton replied that “I tend to do a %listf THATDB@,db — which gives not only the names of the sets in order of their filenames, but also other useful info.”

Tips for system upgrading

With new releases flying for HP 3000s, more users are undertaking a complete system upgrade for the first time. Bonnie Roman said she’d just drawn the task to get to MPE/iX 6.0 and asked for collective advice. Joe Geiser answered with strategy after an excellent checklist was posted by Thomas Genute: Geiser said:

“1. Read the Communicator. Find out what’s changed. The Communicator is the HP 3000 bible for changes between operating systems. Read the technical articles after reading the summaries, for those areas you have a vested interest.

“2. Make sure you have an SLT from your current installation, a printed copy of your configuration (run SYSINFO in the TELESUP account — and print it out). It’s always good to have this around anyway, especially in your disaster recovery manual. Also, print and keep a copy of your network configuration from NMMGR.

“3. Make two backups, both with the ;DIRECTORY option. The SLT and the backup will be immensely helpful if you have to backdate the system. Why two? If one backup was written on media that might be defective or about to become defective, you have a spare. If you have time, run VSTORE on both - just in case!

“4. Schedule this on a weekend. If you’ve never done this before, you will need time. It might go very smoothly, consider yourself lucky.

“5. Read the “READ ME FIRST” Manual that came with your 6.0 tapes before doing anything! It contains vital information on which products may have a slight bug in them (such as FTP). If you are reliant on any of these items which may have problems, put off the upgrade.

“6. Follow directions! If the manual says “Do this,” then do it. Don’t take shortcuts because you will almost certainly FUBAR the upgrade.

“7. I’ve seen people perform an INSTALL instead of an UPDATE. For those familiar with the older MPE/V boxes, an INSTALL is the same as a RELOAD — it wipes the discs clean. UPDATE.

“8. If you have access to the Response Center, be sure to have the number and your system handle ready. If you run into trouble, use them! If you don’t have access to the RC, ask your accounting folks to give you two blank Purchase Orders to use just in case, then use the RC. Yes, you will be billed Time and Materials, but if you are sitting with a dead machine, having the authority to incur T&M is better than having a dead system on Monday, or even worse, a really FUBARed system because you tried to correct something, and made a mistake.”

Richard Gambrell added more tips: “If you use user volumes, each of them and the system must be specified in the store command for the directory.

“Use volutil and dstat to get a copy of the disk configuration. Use buldacct to make job streams to recreate your user account and UDC structure (you may need to keep these somewhere “special” since they contain all the passwords).

“If you use third party system utilities, like MPEX, Formation, or even Qedit, be sure you know how to reinstall the hooks into the system for these products. Take a look at the back-out part in the update documentation to see what else you might need.

“Finally, when you do the update, be sure to login manager.sys from as “clean” and “standard” an environment as possible. Little things like file equates, changes to HPPATH, and UDCs with the same names as system commands can cause trouble.”

Moving databases to disk

Robert Rein asked if there was a Y2K compliant version of the contributed program DB2DISK, which copies databases to disk. While the latest version of this program is available from Interex, several posters noted other methods without membership fees. Chris Bartram said “the DBCOPY program (available in the public domain software area of www.3kassociates.com) will extract datasets to disk files and vice versa. It also copies database to database, and allows selective extracts (i.e. only extract records where key xxx=whatever). It’s also pretty fast, as it’ll automatically use output-deferred mode when copying to another database if it can obtain the proper access.”

Paul Christidis noted that “Query’s save command makes a pretty good ‘DB2DISK’ utility (followed by a delete if needed)” and added that his own favorite was using MPEX and a command file he supplied if dealing with a dataset record size that is less than 255 characters.

Oracle on the outs?

The longest posting thread of the month came from a discussion of Oracle’s shortcomings and track record on the HP 3000. It began when Jeff Kell, the list’s curator, noted that he had decided that “it’s a marketing ploy (well, that part is obvious), but the number of Oracle, SAP, Baan, etc., consultants has reached critical mass such that job security and financial rewards are a huge contributing factor to keeping these things cryptic, complex, and labor-intensive. If there was indeed a 3000 solution a lot of people would be out of a job.” Michel Gauthier, one with direct Oracle experience on the 3000, believed Oracle makes the 3000 better, noting “I would also add that Oracle is probably the best friend MPE could have. The longer Oracle’s RDBMS will be ported on MPE, the longer MPE will live.”