MQSeries.net :: View topic

twegmann · Posted: Wed Jul 02, 2003 6:26 am Post subject:

Here is a general sort of question regarding best practices on modeling/designing a message format.

Some background: We need to integrate some promotion data between applications. The promotion data consists of a promo code and some attributes like name, start date, end date, etc. As well, there are account codes & product line codes that are part of the promo definition (and each of these has some associated attributes like budget & allocation percents). The number of accounts and prod lines varies for each promo. Each promo has a minimum of 1 account & one product line, but could have multiple accounts and multiple product lines. Most of the existing promos only have a few accounts & a few product lines, but there are some oddball existing promos that have 10 accounts & 70 product lines. The business folks tell us that theoretically, they could define a promo with any number of accounts & any number of product lines, though they can't imagine every really exceeding 50 accounts & 150 product lines.

Also, xml is not an option for this particular message. XML would handle the repeating account & product line info quite nicely, but alas, can't use it here.

So, my question is, should I create one big fixed message format allowing for all the potential repeating account & product line codes (many of which would be blank for a given promo), or, should I create a series of header detail messages (the header containing the promo code & attributes, each detail message containing either an account or product line) ?

I have read some theory that says each message should be a full and complete transaction (which in this case would transport a lot of blank bytes needlessly) versus splitting the transaction across mulitple messages and then having to have a way to recognize when all the pieces have arrived to avoid processing only half the transaction.

Are there any guidelines or best practices anyone can recommend on this ? Thanks.