1) The wrapper itself is not encumbered (unless some patent issues applied)

2) Users with a legal copy of Skype on one platform can apply the wrapper themselves (though they cannot redistribute the result)

You can even imagine distributing (via, say, github) a kit consisting of the wrapper and a tool that applies that wrapper to a user-supplied Skype binary. All legal as far as I know (IANAL etc.)

But i have no idea what it IS then

See: http://en.wikipedia.org/wiki/Clean_room_design

What's required (for purposes of courtroom defense) is a clearly documented procedure in which those who have access to the source being reimplemented (the spec writers and testers) NOT be the coders, and vice versa.

Another example is the SMB/CIFS reimplementation in the case of Samba. Here, because the protocol was an over-the-wire system, analyzing network packets (not covered by source analysis proscriptions) was sufficient to create a Free Software implementation.

Software patents are intentionally vague, so I doubt anybody could write a working client based on a patent alone (even if whole protocol was claimed to be described).

The guys who reverse engineered Exchange Server were acquired by Cisco.

I don't understand why the Skype case is 'evil'. distributing the binaries may violate terms of service, but it isn't illegal in large parts of the world.

Unless your home PC is an IBM PC-AT it's the result of reverse engineering.

This isn't true. Nothing stops you from analyzing and implementing their protocol. This project, however, is only the first part.

Their introduction there can be rewritten as "Open source doesn't just mean the source is open. We claim there are also implications about the means of distribution."

It's as though someone was telling people "free beer" doesn't just mean beer that's free, it also needs to be in glass not cans - free beer in cans needs to be called something else…

I think there's a lot of "open source" software out there that is not "open source(tm)". I'm not sure "the open source movement" are doing themselves any favors by aggressively defending their particular definition of a descriptive phrase that is widely known and easy to assume you understand.

(But I also understand the need for specific technical meanings in the software field, and that phrases like "open source" end up being part of the field's jargon, and have far more specific technical meanings than the "plain english" deconstruction of the words indicate. I just think the "open source" case has backed itself into a corner where people assume a plain english interpretation of the phrase, then have the "open source movement" people need to jump on them and say "no, you're wrong". As displayed upthread.)

I understand the opposing view, which claims it's natural to mean "look but don't touch" by "open". I'll grant that it can be a plain english interpretation. But it's an interpretation that violates the instinct of the observer.