My understanding is that it does not matter any more -- use what works for you and your human visitors. I have structure that reads somewhat like example.com/number/number/number and G is able to rank it just great without any problem.

I try to avoid IDs in URLs wherever possible, since I see that as an undesirable technical limitation (albeit one which can't be avoided in some situations). URLs are part of a user's interface into your site, and they should reflect that. No-one is going to remember an ID number.

I would also avoid having an ID number as the first part of the path, since that removes the possibility of users navigating even to the category level via URLs. It's much better usability if users can navigate to www.example.com/category/subcategory/ even if they can't reach the final destination without knowing an ID.

The other consideration is visual recognition of the URL: i.e. a user can know what to expect on the page, based on seeing the URL alone. Again, this is usually read left to right, so I would put the 'worst' element (the ID) last.

S.E.O. aside, how web pages are linked together (structure) is part and parcel of usability and navigation. All of these should be addressed togther before and as a structure is being developed in the design cycle. A short URL maybe advantageous to users when emailing, but a longer "deeper" structure for navigation is more important.

When considering usability as part of a link structure, one should also consider URL length. From a user perspective, "How long will this URL be when I email it to someone?", is important. What would you think if you got an email with a URL broken up on 3 or 4 lines? My usual response is to ignore them ,unless imperative. It is a usability nightmare having to copy and paste all 3 lines into an address bar in my opinion.

One could argue this usability issue is more to do with email that site structure. But it does exist and does effect a lot of websites. Just check your own 404 listings for the full 1/2 URL in a email story. More to the point, how can this very real problem be more easily addressed? Is it with 100 emails to email client writers asking for a solution, or is it to build sites with shorter URL's?

We should also consider that good navigation is part of usability. Being able to get where you want to go next easily is important to a user. A frustrated user will more often click off your site than spend 3 minutes trying to find a meaningful text link to click.

Being able to understand where you are on a site is also an important part of this. Although this can and should be achieved with good meaningful headings and text on a page. A deep link structure, like those of web directories (DMOZ) should also be considered when appropriate. Having a 5 deep link structure at the top of a page tells the user a lot.

With all this in mind I would say when you are structuring a site on a disk that is more than a few levels deep be careful and perhaps think again. When you are structuring your pages from a navigation perspective, having a large directory type structure can be very useful.

So, the emailed URL might be www.site.co.uk/york/ but, the navigation structure on that same page might be Home > Europe > United Kingdom > England > North Yorkshire > York

Marketing Sherpa has a study that shows that a short URL appearing in a SERP below a long one tends to get the clicks. (I have only seen the free summary, not the methodology in the full article).

Keep the directory depth as short as possible.

My preferred choice in dealing with your specific taxonomy might be...

example.com/country/region/city/

It would be natural to follow the breadcrumb trail of the taxonomy in this particular instance. No IDs unless you "absolutely have to" in this scenario. Those IDs may cause more harm than good.

And now that I have "extensionless" available to me on Windows, I'm choosing that path, pun intended.

example.com/country/region/city

No More Extensions
I'm through with them and good riddance!
[webmasterworld.com...]

Whatever you do, make sure that there is only "one entry point" for the final destination. This is where numbers or some other identifier come into play...

example.com/90210-1234

I might look at using zip codes and/or some other unique local indentifier in the URI string if possible. It gets tricky when attempting to make "one entry point" for the destination page when you have a natural "drill down" taxonomy.

Look at the Wiki for clues on how you might work this out.

Shorter URIs have always proven to be more user friendly and intuitive.