With all respect to Mr. Lynch, he is but one more voice declaring some or other form of technology is "dead."  Such obits have appeared for years in the computer domain - COBOL is dead, the mainframe is dead, IBM is dead (killed by Microsoft), Microsoft is dead (killed by Google).  Usually such pronouncements are sourced in a person or company trying to sell the very technology that is supposed to supplant these "dead" technologies and tools.  It seems the thought is that if the mantra is repeated often enough, it will come true and the sourcing company will reap huge benefits from this changeover.  The main drive seems to be to get business leadership to feel that they are falling behind and cannot possibly compete if they are using the "dead" technology.  They need to get on the bandwagon and scrap everything that is currently working in favor of these new, improved means of doing what their legacy systems are often already doing quite well.

Reality has been a great disappointment to these doomsayers.  Mainframes continue to be purchased and IBM makes a great profit from them.  COBOL continues to be used in millions of lines of business code - even some new code.  Assembler, RPG and FORTRAN continue in use, even in the age of Ajax, Java and XML.  Some technologies have, indeed, died - or very nearly so.  But, even in these cases, the particular technologies are dead only because a parallel (not necessarily a new generation) of products have overwhelmed them.  Think of the Betamax, overcome by the still-useful VHS; or Lotus 1-2-3 and Word Perfect, overtaken by Microsoft Excel and Word, respectively; or vinyl record albums (though even these are making a comeback for some audiophiles) replaced by 8-tracks (which did, thankfully, die), cassettes and now CDs, DVDs and mpg files.

There are some great applications that lend themselves well to SaaS - such as project management, geographical mapping, etc.  There are many applications that companies will feel much more secure leaving in-house and limited to their own desktops. 

Anyway, the main point is, it is great to have new generations of technology and they will eventually take a large portion of the market in their areas.  However, older technologies will remain in play for a long time to come, as businesses wisely refuse to spend money simply to get the latest, shiniest bauble, while the old diamonds continue to work just fine with a little polish.  Desktops are not dead, nor will they be for a long time.  They will simply be used in conjunction with the newer SaaS apps where it is most feasible and provides the greatest improvement in service or cost.