This is an abhorrent practice. It can be misinterpreted in so many ways. ±.25 from what? Can two holes that are .5 apart have the same tolerance as those 200 apart?
At the very least, specify datums from which to measure. Using a blanket GDT surface profile callout with specific datums is much less confusin (though still likely to cause trouble).
Applying a general tolerance to solid models is a practice of growing popularity. Simple +/- tolerances don't work for the reason stated by Roland. Where is the tolerance applied from? Your part is basically undefined.
There are currently three options to fully define a part from the solid model (as of ASME Y14.5-2009):
1. Add all dimensions in the model space that would've been added to a drawing (full PMI - ASME Y14.41-2012 covers requirements for this).
2a. Declare the model as Basic, and apply one Profile feature control frame to all surfaces referencing stated A, B and C datums.
2b. Declare the model as Basic, and apply one Profile feature control frame with ALL OVER modifier. I talked about this here some time ago: http://www.fcsuper.com/swblog/?p=213
The safest bet is #1, but time savings isn't quite as good as #2a or #2b.
#2a is fairly clean, but relies on knowing what is A, B, and C datum on every part. (Three datums are not always required/available, nor is it always possible to apply these to design intent.)
#2b is cleanest and most consistent, but you'll have to work out with your vendors on how you will judge their final products directly against a solid model on a computer screen. 3D scanners are getting much more accurate these days, but the better ones are still pretty costly. You can use the same methods as you do for #2a (CMM or OC), but some dissemination of data will have to occur for undimensioned features.
Edit: These are all Model Based Definition methods (that is, defining the product from the model)
Message was edited by: Matthew Lorono