First of all, a convenient approach is to "publish" the long version as a technical report or as a preprint (e.g., on arXiv). In this way, it is always possible to refer to missing parts in the shorter version.
If there are only a few pages too many, then it's probably possible to "tweak" the paper to fit within the page limit: scale the pictures, inline the equations, use only acronyms of the conference in the bibliography (e.g., in POPL, instead of "in Proceedings of the 40th Annual ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages), transform subsubsections into paragraphs, rephrase some paragraphs, move the proofs into the appendix, etc.
However, if you have too many pages, and you actually need to cut some content, then it might probably be better to directly cut an entire section rather than some bits and pieces in each section. For instance, I've rejected some papers because "due to the lack of space", the authors didn't provide any illustration of some quite complex notions, making the paper not understandable. Similarly, I've rejected a paper that was addressing many different points, but never in depth, due to the lack of space ... so it was quite hard to see the contribution (it was not a survey paper).
If you can't find a proper section to cut, i.e., you make a single contribution and you need 15 pages to explain it, then the venue is probably not suited, and you might rather look for a venue with a larger page limit.
Basically, if you have a 15 pages paper and a 8 pages limit, write two papers!