At the time this looked like a negotiating tactic: using the court the get IBM's senior people to pay attention to an internal problem. To my astonishment, however, IBM choose to deny SCO's claim, embarking on a costly, and high risk, legal process rather than settlement negotiations.
I don't pretend to understand American civil law, but it seems to me that the basic SCO claim is an absolute no brainer: one line of AT&T code in Linux makes the case and the original 390 port done in Germany and released through SuSe was chock full of things that seem to have benefited rather directly from work done under the AT&T shelter by engineers from Sun, NCR, Motorola, and AT&T when the BSD/SysV union that gave rise to Solaris was co-developed for PowerPC.
Unfortunately IBM's decision to circle the wagons led to a rapid escalation in claims on both sides. Thus SCO hired a well known east coast law firm on contingency plus fees and expenses, and the subsequent filings raised the stakes considerably by alleging that IBM's senior management was complicit in this process, issuing uninformed commentaries on Unix history, and then demanding licensing fees from the general Linux community in a rather naive attempt to establish a fair market value for the formerly protected code and ideas.
Those mistakes allowed IBM to rally the Linux community against SCO, empowered anti-SCO sites like groklaw, and created an intensely negative emotional atmosphere in which hardly anyone noticed that IBM's actual legal strategy consists of a large red herring (the dynix file system code; something SCO's lawyers bought hook, line, and sinker) coupled with alternately stonewalling, nitpicking, and drowning SCO in paper.
Four years later some clarity is emerging. SCO's lawyers are responding to IBM's strategy by broadening their case from code to the ideas embedded in that code. At the nit picking level, for example, if there are four equally good ways to handle a high/low switch and IBM's original code used the same one AT&T did, it may be reasonable to believe that the guy who nominally wrote the code asked the AIX maintainer in the next cubicle how AT&T handled it. At the more general level there may be a number of reasonable ways to do kernel table management, but if only the one invented at AT&T and uniquely embedded in the licensed AT&T code appears in IBM's product and results in other significant simplifications that also appear in the kernel, then the balance of probability falls on SCO's side.
Indeed if SCO can find even a couple of compelling examples like this, particularly at the level of significant design and coding optimizations, the likelihood of independent invention starts to approach zero.
The trap for IBM here is that SCO can argue that copying at the level of design ideas indicates management complicity because, at the very least, the product development managers would have had to sign off on both design and code. And if the court buys the argument, this means that IBM's loss will be far more consequential than was ever contemplated in the original filings.
Speaking personally I see SCO as the aggrieved party here - but I think their real complaint today should be as much against their own lawyers as against IBM. The basic case, however, was and is a no brainer, and should have been settled without the damage to SCO's reputation and ability to do business we've seen so far. Thus the bottom line is simple: if SCO survives long enough to make it in court, IBM will lose.
As I've said before, there is a natural basis for a settlement here under which SCO would drop the case and IBM would pay SCO to release all of its remaining rights in Unix code to the public -allowing IBM to leave the field in triumph after having achieved something of value to the Linux community.
Most people, of course, deny SCO has a case and see no value in open sourcing the AT&T product set. Now, if you're in that camp, let me ask you two questions: first, from whom are you getting your information? and, secondly, if you think open sourcing AT&T's code now would produce no change in Linux or any other major Unix variant, I'd guess that you'd be right; but have you considered that you're making SCO's case for them?