[lwf]

> every redirect has a cost:

If you are actually just keying your content lookup on the ID and don't redirect the user, what's the performance problem?

And use rel=canonical so search engines do the right thing.

[franze]

no

simplified google works like this

discovery (queue) -(quality check)-> crawling(optional) -QC-> indexing

google does not "follow" canonicals, but whenever google discovers (during crawling) a canonical it pushes it back to the discovery queue -> needs to crawl again -> needs to figure out indexing

canonical is an indexing directive

so basically there are two quality checks before google can actually apply the indexing directive after it has discovered the canonical during crawling. also you can never be sure when - if ever - it will fetch the canonical URL or choose to canonical it.

for small sites this is not a big an issue (you will have internal duplicate pages for google for an unknown amount of time, but at one point they will probably be canoncalized). for big sites with millions and millions of URLs this is a big issue. basically in your example is the worst case: URL rule 6 (least important) breaks rule nr 1. then why do it at all

additionally to communicate different URLs to the users (based on the way which they came to your site) which is just bad UX.

don't do it.