As far as I know .MDL are used as a back up, and the main difference between .PY? & .MDL is that the MDL does not stores security.

As the .PYI stores security, the model can be run in batch, otherwise the model would prompt for the password to the database and would be waiting it.

That's all. Hope this helps

Einyel
Functional&Technical Cognos Consultant

Also PYx runs a bit quicker, as it does not have to be compiled. Speed differnce is very little however.

Bruce

mdl's do store security in them, they just won't embed a database password. The main differences between mdl's and pyi's are as follows:

Mdl's are smaller as they contain JUST the information required to build the cube. They are also the 'supported' way to upgrade from version to version as they are portable.

Pyi's can grow to be quite large as they are binary files and can become 'unfragmented' and tehy tend to store a lot of legacy information that you can no longer access. The upside to these is that they can store signon information and will run SLIGHTLY faster.

flex

An additional comment to Flex13.
When your model is developed using PYx, before using it in test or production, save it as a MDL file. Then open the MDL and save it back to PYx , it remove the fragments and the PYx size decrease.

Thanks all for the replies.  In summary of what I've read:

PYIs can fragment but have the ability to store signon information and run slightly faster runtime than MDLs.  Fragmentation can be reversed by saving to MDL and then resaving as a PYIs.

MDLs are slightly smaller, slightly slower, don't fragment, can't store signon information but can be used when upgrading cube models from one version to another.

Thanks again. Maybe we need a poll of how many people use either PYI or MDLs regularly as their main storage format.  :)

My Poll answer:
MDL's - Main
PYI's - Storage/backup

CP  

I always did the exact opposite of CP, so I guess either 1) I did not FTFM, or 2) different people have different needs.

Another plus of MDL is that you can read the file directly in a text editor. I had to use PYx to use sign-ons & run unattended cube updates.

Bruce

I'm with Bruce, and probably for the same reasons.
I use PYx for the regular builds, and then a save to MDL and then back to PYx to 'compress' large models that have levels which grow over time (document numbers).
With the signons, it also allows other (trusted) users to kick off cube builds when you're not around/available.
lex

Correct, MDL's should be the backup/storage mechanism, where PY's should be the runtime models.

flex

I create and edit the .mdl's (Cognos claims they're incorruptible, and so far none have, while plenty of .pyi's have been), then save as .pyi's, and put the database logon info into the pyi's, then save again, for the macros to run. I never do any real work (editing) on the pyi's, they absolutely get corrupted.

And as I found today with a REPOS-E-NOMEM error on a pyi that'd grown to 290MB, it's extremely useful to have a mdl as a back-up.

With the data on the lowest level removed, the mdl was a svelte 6MB.

Had I just had a copy of the pyi, chances are I'd have had to create the model again from scratch.

lex

It shows you need both types kicking around the hard drive.

Bruce