The only risk in this is exposing cookie-based accounts that have a fair amount of rep associated with them as being possibly abandoned. If the last activity is more than a few months ago, someone's lost their cookie and probably won't realize they're disenfranchised from those contributions for months, years, or possibly ever.
Someone with less than noble intentions could then:
- Find unregistered accounts from years ago that still have a gravatar hash
- Use a rainbow table to get the email
- Request a merge and possibly profit, since they'd basically know the only non-public thing about the account. (Geez, I think I was in an Airport in Belgium when I wrote that answer that got 40 votes, can't remember)
- Probably uber-rare these days, I have to dig a bit, but still a concern. I've caught what I think is possible attempted abuse of auto-merging; a process which we've made no secret about in the past. That's what peaked my eyebrow.
You can happen upon these now opportunistically, but there's no easy way to query for them. Adding that type gives you this ability.
Beyond that, I have no real misgivings about including it in future dumps (which would basically let you backfill all previous dumps to some extent). What you won't get is any means of knowing if an account we now say is registered was unregistered 3 years ago. If it's unregistered today, it was unregistered at any time in the past (you can't go from registered to unregistered).
There are very interesting use cases here:
- Moderators looking to examine how different sets of people actually use their sites
- Researchers interested in many aspects surrounding anonymous participation
- Researchers working on things like tag prediction (a very common thesis these days) wanting to split the user base even more while looking at what their implementation picked vs. what a user actually went with
- ... there are uses for this in the dump, let's just leave it at that before I start wandering
I have to think about it a bit and make sure I'm not missing something horrendous that might happen if we included this. If I come up empty, I'll recommend internally that we do it.