bullet proof arguments for switching to Manifold in general GIS workflow
That is a compliment for sure, but in truth our objective is more modest, for Manifold to earn a place in your GIS toolbox regardless of which tool you prefer for your primary GIS.
I think that's especially true for Release 9, which has just very recently been issued and is evolving very rapidly. Some people can already use 9 as their primary GIS but others will be using it to augment their other tools. Over the weeks and months ahead as 9 gets more and more features there will be less need to turn to other tools to do what 9 cannot, of course, but for now there are key areas yet to be integrated into 9. Web serving is one such area.
Release 8 has a spiffy web server, IMS, but 9 as yet does not. 9 will get a web server too, significantly better than that in 8 we all expect, and anybody who has a 9 license today will get that as part of a free update to their Release 9 Universal license. Once that web server comes out it is likely that versions of 9 licenses (as was the case with Release 8) will become available at "personal" or "professional" or whatever levels that do not include the web server as Universal licensing does. But for now, the only 9 licenses that are being sold are all Universal licenses so if you get 9 today when the web server comes out that will be part of your license as well.
To comment on your other questions (other folks on this forum who know more might want to pitch in...):
* How does Manifold compare with GeoServer/MapServer in delivering WMS services, especially under heavy load? Have any benchmarking studies been done?
MapServer is a fine product. For various technical reasons I expect anything in 9 would be the same or faster. There's no reason to believe that 9 would in any way be slower, whether 9 is running off native 9 project storage or 9 is running off data sources like Oracle or PostgreSQL.
* Since GeoServer is pure Java, I think it would support 64bit architecture/parallel processing to the extent that Java natively supports it. Is there a way to incorporate Manifold (and other GPU/Parallel stuff) into the web-publishing/accessing workflow to make it more efficient.
I think for now it would be most effective to let Manifold turn the crank on 9 features for the next few months to see what emerges for handling web serving.
You probably didn't mean it that way, but just in case for others who might not understand, it is worth mentioning that using Java in itself doesn't add parallel processing to anything. You have to write parallelized code to make what you do parallel, not an easy thing for most people to hand-code regardless of the language they use.
9, in contrast, is automatically parallel and whatever you write in SQL in 9 will automatically be parallelized. You don't have to code your own parallelized routines... just write ordinary SQL and 9 will make it parallel.
That brings up a speculative case for 9 in performance as compared to environments which are not automatically parallel. Sure, if all you are doing is serving tiles for a particular area of interest, well, that's not rocket science and whatever the WMS server you use does by default is probably going to work OK. But many people operate web servers where they want the server to alter what it serves based on user criteria.
They want servers, in effect, to create results that are based on queries, which in some cases might be queries that are way more complex than just showing those dots that represent houses for sale in Hermosa Beach, CA, that cost between $800K and $3 million. They might want the end result of several sophisticated operations, like topology overlays and spatial joins, all put together.
Such work is all about sophisticated DBMS queries, and usually those are best done in spatial SQL, not hand-coded in C++ or Java, since there is a really big amount of DBMS supporting work done for you automatically in SQL. To the extent you can use a highly optimized, fully parallel, spatial SQL instead of hand-coding something, the better performance you are likely to get as well.
It's been well over a year since the automatic optimizer within the Radian query engine used in 9 has started producing faster results in spatial SQL than most human programmers can hand-code in C++. These days you can use spatial SQL in 9 and be fairly confident no group of humans is going to be able to get better speed than your SQL by hand-coding. It's also highly unlikely that anybody's first or second cut at such things is going to be as totally bulletproof, or as maintainable or as updatable as the SQL result, either.
I guess the bottom line answer to your question is that using 9 as a web server right now is not yet an option, but that there is every reason to believe that what 9 does eventually deliver for that will be a good option.
Ah... one last thing. The nice thing about having quality software like MapServer as a point of comparison is that nothing stimulates immediate improvements within 9 as saying, "hey... this software over here works faster...".
The recent exploration of 9 speed with contours is a good example: any case where somebody finds any instance where something else is faster is going to get an immediate response. Such edge cases are really super for finding places to improve, and Manifold makes such improvements in days, not weeks or months.
With contours, I think today's build or tomorrow's will add tweaks that increase speed dramatically so the one or two examples where ESRI was found to do contours faster are now set right, so that Manifold runs significantly faster. Those improvements happened within two days after we first received the sample data required to run tests. You'll get that same response with any comparisons of 9 to MapServer as well. If 9 ain't at least as good or faster, Manifold will work night and day to get that improved right away.