(it's NOT a front end/back end database).
see thread181-30072
I do not recall that the total discussion of the arrangement is in this thred, but the code to do the auto update is here.
The arrangement which is the target of the discussion (for MY multiuser maintenance) is:
Each User has the FE installed locally.
There is a MASTER (FE) which is where ALL approved / finalized updates are placed.
I maintain a Working copy (of the FE) on the 'db-admin' machine (or server), which is NOT the Master.
Each proposed change is reviewed and approved by the User group. This group 'meets', discusses and votes on each item, usually in the forum of e-mail. I (db-admin) implement all approved changes in the Working copy and do some testing. When I think it is correct, it is presented to a seleted group for further testing/commentary/re-pair.
When 'testing' is complete, the changes are implemented to the Master copy. The routines (see referenced thread) automatically 'update' the users FE copies the next time they start the program.
There are some other 'issues' which should be addressed in the use of multiuser dbs. I will not attempt to 'resolve' them here, but just list some topics. I and others have commented on most of these extensively in these fora, and some soloutions are presented for those who care to use hte search capabilities and winnow the wheat/chaff stuff.
First, the referenced thread will not work with the MDE format, so if there is an administrative reason for the use of that (.MDE) format at least the update process cannot be done (at least not in a piece-meal sense). You could use a different process to detect a NEW .MDE and copy it wholesale to the user systems.
A:[tab]I maintain a form / table in the db for users to report issues problems to me through the app itself.
B:[tab]ALL useful dbs should (MUST?) implement security. Without the security protocalls, many users could easily generate the trivial procedures to destroy the database. Som (actually MANY) of the ways this could be done are close enough to legitimate work that ic could be done mistakenly. Placing the db into the .MDE format will not prevent some / many of these.
C:[tab]If a db is at all useful, it must be backed up. As 'they' say "sttuuuffff" happens. And -at some point it WILL! Many network backup schemes rely on Win (O.S.) for the actual file copy. Win WILL not copy a file while it is in USE. So, you need to be quite sure that the 'net cops' are actually accomplishing the back-up. Ask them -in 'writting'- several questions
[tab][tab]1. Do they review the back up log on a daily basis
[tab][tab]2. do they notify each "owner" of EVERY occurance of a filure to back up hte files htey (the owners) are responsible for?
[tab][tab]3. Can (and WILL) they 'disconnect' all users (of the files YOU are responsible for) before doing a backup?
[tab][tab]4. Can YOU review the backup log re YOUR files at least occassionally?
If you are not completly satisified with ALL of the above - go directly (and in writting) to your supervisor. Do not "STOP" an informing the super of your concerns, but go straight to the point that he takes decisive action to resolve these concerns.
D:[tab]Multiuser databases often need to review the sequence of events in a process. This is ONLY possible if the individual system clocke are synchronized or every 'transaction' uses the server clock. I (at least so far) have found no pratical soloution involving the latter, however syncing the individual systems is a routine network process, so -AGAIN- address this along the lines of the backup. Formal correspondence to the 'net cop' of choice through formal completion / acknowledgement.
E:[tab]In realistically maintainong a multiuser db, some issues of 'who-shot-john' (and WHEN did the shooting occur) will arise. The 'reality' game of CLUE. To provide the evidence, you MUST have a transaction log, and it MUST be complete and accurate (back to ye olde time stamps'!). (I have written faq181-291 which can be used for BOUND forms in Ms. A. whichcovers this.)
Most 'transaction logs' quickly grow to be larger than the databases. You will need to have a process which 'archives' the transaction log records to an inactive storage on a regular (and perhaps frequent) basis. Ideally, transaction logs should be on DIFFERENT servers. Transaction logs not are useable as a method of 'reconstructing' the data up to the point of a 'crash' - but they will not be useful if you wait for the crash to occur before writting (and testing) the procedures which do the re-construction. Again, from MY perspective, this should be an important issue and receive priority attention (and 'funding') from the supervisor - with appropiate (and FORMAL) documentation.
MichaelRed
m.red@att.net
Searching for employment in all the wrong places