Are you saying that GPL holds off forks better than BSD/MIT? If so, by what mechanism? Of all the pro-GPL things I've heard, I've never heard that one before, so I'm quite interested to hear this fleshed out into something I can follow!
The original license from Bell Labs for Unix wasn't BSD/MIT, but it had essentially the same effect (and in fact, BSD is a down-stream license of it): you could get the source code for Unix and do anything you wanted with it.
Most people started with the BSD distribution (because networking), many later incorporated System V things from Bell. They then added their own changes and released the result under their own proprietary license. "They" being Sun Microsystems, IBM, HP, Silicon Graphics, and a host of smaller organizations including NeXT. Each of these had some really neat feature, like IBM's Journaled File System, SiliG's graphics, and Sun's hardware and the fact that they hired Berkeley's CSRG people. But they were all as incompatible as they could be made to be and still be Unixy. They were also business rivals and fought each other constantly. (I was part of IBM's effort with the "Open Software Foundation"'s DCE (https://en.wikipedia.org/wiki/Distributed_Computing_Environm...), which attempted to fight off Sun's lead with NFS.) This was at a time when Microsoft was pushing Windows on smaller machines and those smaller machines were getting powerful enough to be a threat in the Unix "workstation" market. As a result, Microsoft won the Unix wars.
Linux, in contrast, is under the GPL, as are most of its basic user-space programs. You can modify the code to add some feature, but you have to distribute the changes if you distribute the whole thing. As a result, your feature gets incorporated into the mainline kernel, or you fork Linux. Forked Linuxes don't get a lot of traction, though, in part because nobody has a serious hardware advantage, but mostly because the mainline quickly grows a corresponding feature. It's very hard to lock your customers in to your own distribution.
