In-Reply-To: <199507051818.NAA19699@riverside.mr.net>
Message-ID: 
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On Wed, 5 Jul 1995, Jeremy Allaire wrote:

> David Oliver writes some interesting though
> mostly uncompelling arguments about the
Jeremy, You are showing contempt prior to investigation.  Have you
actually LOADED a copy of Linux onto your PC?  Try it - it'll cost you
under $50, take about 8 hours, and you can easily go back to your windows
environment - I have been using both on my PC's for years.
I reccomend Slackware (most flexible) or Plug'n'Play (easiest to install).

> need and utility of UNIX as a Web Server
> platform.  Here are my responses:
> 	
> 	<< 	there would be no Web, Jeremy.  And, no
> 	Internet.  And, no client-server.  >>>
> 
> No; the Web is popular today because a number
> of developers understood that the Internet's future
> was in graphical desktop clients -- mostly Mac
Actually, we (of the internet infrastructure support groups)
knew that back in 1983.  That's why, when the Mac came out
in 1984(?) Bill Joy was showing off his Sun 1, complete with 68010
CPU in our computer room on a trial-demo.  Something we picked
up from the boys at PARC-place.

> and Windows, which were the driving forces in
By the time Bill Gates had a version of Windows that could run
on a platform that could support it (Windows 386 or Windows 3.0)
Unix was on it's FOURTH release of X11 (R4).  In fact, Gates was
inspired to put many of the features of X11 into Windows 3.0.

For a while, there were several lawsuits because HP allegedly was so
much like windows that you could only tell when you realized that
only the active task-window was able to do anything on the "Real" windows.

> computing, particularly Windows.  Had no one taken
> the time and made the effort to make the Internet
> a non-UNIX world, the Web would be no where but
> a bunch of academic and R&D babble.

I am personally responsible for putting out the first commercial software 
directed at making the internet available to Windows users (SoftWin - from
SoftTronics Corporation).  We made it as simple as possible to download the
core modules, enter a few "phone numbers" and have the user up and running.
I have since left that company - but they were the pioneers in that 
technology.

I was also responsible for getting the internet into the Business 
community.  I spoke the the Director of Frame-Relay sales at MCI in 
Colorado.  I convinced him that he could easily sell frame relay by 
supporting IP routers and supporting TCP/IP - I also pointed out that 
some businesses would be very interested in accessing the databases of 
the National Science Foundation.  As a result, MCI practically gave away
their internet access to the NSF in exchange for permission to let commercial
users "come and visit".  That, followed by a series of strategic alliances
has been the foundation of the exponential growth of the internet.

Even as these late-breaking developments were taking place, Linus 
Torvaldis had already developed a significant following among the 
"hobbyist" computer users (the ones who wrote all that PC Share-Ware).
The had soured on paying $500 in software royalties for their "hobby 
machines" to be developers.  Linux was free and supported all of the
unix applications - including X11 windows.

Since 1982, Microsoft has constantly been in a frantic race to stay one 
step behind UNIX, and appear to make the comprimises (monotasking, single 
user, no security) look like benifits (cheaper, easier to use, ...)

AT&T certainly didn't help UNIX much.  Their insistance on a price floor 
of $700/copy + $100/user was rediculous.  Even when BSD 2.4 was available 
on a PC XT (needed 128k and 10 Meg Disk drive).  It wasn't until Noorda
(Novell) purchased USL from AT&T that anyone priced Unix for Workstations.
I might point out that SCO unix had already established a very strong 
niche in the PC-Server market.

> 	< 	bloom - free tools, thousands of people
> 	contributing things on their own time,
> 	etc.  >>
> 
> This is partially true.  Yes, thousands of people,
> free tools, etc., helped, but it was rather
> the distributed nature of the system wherein
> anyone could become a publisher -- while the

Actually, anyone could publish anything.  The Unix user community
insisted that code be made available in source code form.  I posted
a binary to a unix group once (for a VAX), and found my mail-box
packed with requests for a version for their machine.  In two days,
I published the source to the comp.sources group.  The source code
was a strategy for converting C arguments into a string, sending the 
string to a server, and having the server convert the string back into C 
arguments and call a specific subroutine then convert the results back 
into a string and send that string back to the client.  I called it a 
"virtual call" (after virtual memory, virtual machines, and virtual disks).
Within a few weeks, Larry Early wrote a lex/yacc parser to do the job via 
a compiler.  A few months later Sun announced the Remote Procedure Call.
Unix was the birthplace of client-server computing.

Until the Winsock specification was implemented - 10 years later, it
wasn't possible to create a practical server on a Dos/Windows client.  
Beyond a few users, the system would collapse because the applications 
always had to be completely self-contained.

arc, zip, and "tsr's" are all applications which emulated unix features 
but had to be reworked to compensate for the monotasking MS-DOS environment.

Microsoft even had a version of UNIX (Xenix), which it abandoned to 
create OS/2.  As a contender in the Unix market, Xenix was not 
spectacular.  As a contender for the Microsoft "next generation 
platform", it was better "out of the box" than the later "better than 
unix" products.  Selling Xenix to SCO may have been Bill Gate's "Waterloo".

> tools have been poor, relative to the quality
Compared to what?  If you want to make comparisons - compare
shareware to shareware and compare commercial to commercial.

> standards set by such applications as Microsoft Office
The first office automation packages were developed on UNIX.  We had
e-mail, document flow, resource sharing, databases, and most of the other
popular features of office back in 1983.  In the evolutionary nature of 
unix, we have adopted new features such as Mime for sending Mpeg, Jpeg, gif,
bitmaps, and postscript.  We have 5 different word processor grade 
typesetting standards (nroff, TeX, sgml, html, X11, postscript).  Because we 
support the same file standards, I can choose from the "freebee" word 
processor, or I can use one of several expensive products like FrameMaker.

> or Lotus SmartSuite, the fact that no one controlled
Again, because we have a set of publicly defined standards, we have 
several interchange formats - some of which are even known to Windows Apps.
Lotus 1-2-3 for Unix is much more powerful than the Windows version.  
FrameMaker makes Microsoft Word seem like a toy.  Even the shareware 
outperforms Microsoft Draw and Paint.  Even the Microsoft Foundation 
classes run better on Unix.

> the production and distribution of content was/is
> the driving force -- not some world of UNIX
> development.

The infrastructure of UNIX demanded that no single vendor control or 
dominate the market.  When alliances formed that threatened to exclude 
one group, they would form another alliance and create their own 
enhancements.  Often the "losers" would "give it away", choosing not to 
spend time/energy on "hyping" a pointless feature.  Slackware Linux 
offers three different "loser" window-managers for free.  You can buy the
"winner" (Motif) for under $200 if you really want it.  You can get the
Andrew and Doc word processors for free, or pay $200 for WordPerfect, with
full graphics capability (faster and more powerful than WP for Windows).


> In fact, the overwhelming majority of Web developers
> are using Mac and Windows tools -- there's a reason,

Actually, the overwhelming majority of Web Servers and Server developers 
are using Linux, Solaris, OSF/1, AIX, and the other 20 sub-brands of Unix.
It wasn't until recently that a vendor could actually call it's product 
Unix without charging a minimum of $700.

> they're simple and don't require knowledge of a C
> shell -- a remnant of the past, I hope.

C shells are very useful at times.  I often wish I could have a few CLI 
windows in MS-Windows.  If you don't like C-Shells, there's always Perl and
utilities written in Tk/Tcl (Makes visual basic look complicated).

> 	< 	Jeremy.  In fact, Linux is free, or available
> 	on CD-ROM for $39 with a complete development
> 	environment - including the best compiler suite
> 	in the world from the Free Software Foundation.>>
> 
> For those familiar with cost-accounting, the above
> may seem a bit silly.  Linux is 'free' just as a
> prisoners lunch is 'free'.

First of all, let's separate the apples and oranges.  To be accurate, 
Linux has no intrinsic royalties.  On the other hand the service, and the 
quality of service, is what you actually pay for.  Because of this, the 
Linux market is a highly competitive market of about 10 vendors who serve 
their costomers with varying degrees of quality.

> Along with Linux comes a life
> of misery at the command line, where users must
> complile every program they want to run, where there is

If you load your raw linux directly off the internet, this is true.  If 
you download the 90 megabytes of the internet (got about 16 hours you 
don't know what to do with?) you get more.  You can buy a CD, with no
book, and print the documentation from the CD for about $20.  You can
pay $60-$80 and get a nice big telephone-book size manual pre-printed.
If you are really lazy you can pay about $400-700 and have someone 
pre-install it for you.  For about $1400 you can get a PC pre-configured 
with Linux, ready to go.

> no formal support, few major commercial desktop applications,
> and certainly no Industry standard plug-and-play (see Win95).

Actually, the default is plug-and-play.  In fact, if figures out most of 
the old dinosaur cards (3c503 and 3c501) that Win95 can't even deal with.
Plug-n-Play on Win95 requires custom software from each hardware vendor.

Realisticly though, Linux does have some limitations on hardware.  Some 
companies contribute their drivers, other let you download them from the 
internet, others are not yet supported or are in alpha (gambling).

> Let's get real, David, I don't see people running to the
> software store begging for a UNIX aisle -- the bottom line
> is that the real cost is in the enormous amount of time
> required to use such an archane and complex system such as UNIX.

Most people don't buy unix software in a store.  Because Unix is a secure 
system (once security is configured) vendors are more willing to upload 
software or upgrades.  Often, you can get software through the same 
sources as your corporation buys them.  I do my linux software shopping 
via the web.

> There was a reason for the Mac and PC/Windows . . . to free
> people from DOS, from having to deal with commands and syntax
> and what is more appropriately the domain of hardware
> and software engineers. 
I appears that you have not yet tried the X11 Windows interface.  I 
reccomend the open-view window manager, and the Xfm file manager.  You 
can drag-n-drop just like you do on a Mac.

> Furthermore, PC-based graphical
> computing was a cognitive revolution, it was an evolution of
> computing tools that enabled a greater and greater share of
> people to use them.  Let's not take a step back.

PC-based graphical computing was "revolution" only in the fits and starts 
it took to have the same techology that was available on UNIX finally 
show up on the MS-Dos based box.

> 	< 	interface tools.  There are literally thousands
> 	of Common Gateway Interface programs and 
> 	scripts available in the public domain (free).
> 	Perl, the language of choice for much work at
> 	Web sites was written first to operate under
> 	Unix - perhaps the first language ever to unify
> 	systems programming and application scripting.>>
> 
> Of course, and most of them are essentially useless outside
> of basic development.

I suppose for you, the web is useless too?  One of the biggest drawbacks 
of Windows(3.1,95, or NT) as a SERVER is that one has to interact with 
the mouse and keyboard attached to the PC.  When the Vax or the Unix 
machines go down, I can take corrective action from my office.  When a 
Windows box goes arry, I have to take a trip up to the computer room
(and miss phone-calls...)

The ideal server doesn't even have a Keyboard or a Screen.  The Crays 
used to have a PC that sent the management instructions.  Today a 
pentium/90 is more powerful than a cray 1.

With Unix/X11, I can manage 50 machines from the same console.  With 
Windows, I have to walk to 3 different floors to take care of them.

>  Those that enable more complex development,
> such as Perl or C integration require just that -- Perl or C
> integration.

When your only dealing with 5-10 hits/hour on a simple web page, You 
really don't need Linux/Unix.  When you are dealing with managing 500 
transactions/seconds across multiple T1 links, you need the power of Perl 
and C.

>  Companies, IS departments, schools, etc., are not
> just about to run out and abondon their knowledge base in favor
> of some system that is more complex, costs more in training,
> and require enormous amounts of time for building any sort of
> sophisticated application.

This sounds exactly like the argument I used to get from the Mainframe 
Managers.  Of course, when it becomes trivial to string together a few 
shell commands and a 20 line PERL script to impliment a multiuser 
client/server application, it suddenly makes sense to use Unix as a server.

> CGI interface tools, database gateways, and other things
> that are currently available are pretty limited -- what's needed
> are tools that don't require ANY knowledge of CGI, that plug

The beauty of CGI is that the Clients don't have to have any knowledge
of the underlying server.  The CGI interface lets the user click a few
choices.

> into a high-performance multi-threaded and multi-processing
> system, that tie into the vast majority of database that
> exist in the world today, and further, that play into the hands
> of the largest community of software developers in the world --
> e.g. Windows developers.  This happens to be the community/market
> that will drive tools development on the Internet, whether you
> like or not.
> 
> 	<< 	1993 through today, we've seen astonishing grow
> 	in the number of tools and applications available
> 	for the Web.  I could argue that even the "PC
> 	revolution" was slower at getting started than
> 	the Web.>>>>
> 
> Yes, in Nov 1993 the Web was introduced to the standard
> computing world -- e.g. PCs and Macs.  That introduction
> helped bring the world of non-UNIX computer users and
> developers to the Internet -- it also created an enormous
> demand for tools; not UNIX tools, PC tools and Mac tools
> that could free these users from having to deal with such
> a difficult system.
> 
> Furthermore, lets take even a _cursury_ look at the Web
> tools market.  Early on, there were many freeware tools,
> and continue to be shareware and freeware tools on all
> platforms.  But lets take a look at where the
> commercial computing market is taking things.
> 
> Servers:  there are over 10 commercially developed
> and supported servers for Windows and NT, 1 for the
> Mac, and 2 for UNIX (Netsite, OpenMarket).
> 
> O'Riley, long the bastion of UNIX gurus, is now out
> pushing NT as the future platform for the Web, 
> shoving, of course, there absolutely fabulous
> NT-based WebSite Web server  -- you can actually
> administer it without having to *touch* a command
> line; no compile involved.  Remote admin, multi-homing,
> runs as a service, etc.  $300 at your local bookstore.
> Try it:  http://website.ora.com.
> 
> There are well over 30 companies building various
> HTML authoring tools, mostly for Windows.
> 
> 	<< 	"blamed" for this growth. >>
> 
> Just as I might thank UNIX for inspiring WindowsNT,
> it mighty also thank WindowsNT from freeing me from
> UNIX.  Likewise, I'll thank UNIX for inspiring the
> Web, and thank the Web community for freeing me
> from UNIX.
> 
> 	< 	languages, new "general solutions" just
> 	as physicists are searching for the 
> 	Grand Unifying Theory.  >>
> 
> Yes, of course they are.  And businesses are looking
> for solutions that can be delivered today for a
> minimum of cost and that are likely to be future
> candidates in the technology landscape.  While some
> mighty put there eggs in the Java basket, others
> mighty simply wait for the days of Cairo -- 
> distributed and secure OLE apps -- which, incidentally
> is well beyond the alpha stages of Sun's technology.
> 
> > First, as Web publishers, we must be asking
> > what we can do to enhance and develop our
> > technologies to be more dynamic (literally,
> > no static pages) and user-driven -- more
> > interactive, not necessarily in the user-to-user
> > framework (though this is a big part of the issue).
> 
> 	<< 	late to the party.  The world has embraced the
> 	WWW, instead of say WAIS or Hyper-G or others,
> 	and the WWW could be seen as "server-centric".
> 	The server-centricity of the WWW makes it 
> 	easy to "distribute" applications power, but
> 	not to "decentralize" it - I think that is
> 	what you are asking for user-to-user.>>>
> 
> This is not my point at all.  My point is that people
> need to build more than hypertext pages; they need to
> build true client-server applications with state
> systems, semaphores, real-time data and graphics
> generation, and a high-degree of user choice
> and interaction -- not clicking hypertext.  All of
> this is quite doable with existing Web protocols.
>   I agree
> with you very strongly that the Web has been embarced
> because it allowed for the easy distribution of such
> applications.  Some common ground?  :-)
> 
> For a demo of what I'm talking about, see a tool
> I developed -- a basic Web conferencing system.
> No code was involved in writing the system, only
> embedded SQL.  It's running on Windows NT.  It
> took me about a week's time.
> 
> http://www.creativeis.com/talk/working.htm
> 
> 
> 	< 	do - either by locking you into one programming
> 	model (Mac) or locking you into a narrow operating system
> 	and class system (PC).  Just as you can not have
> 	simultaneously dictator and democracy, you can not
> 	have programming rigidity and programming freedom.>>
> 
> This is simply not true.  Practically every programming tool
> on UNIX is available on NT.  Fortunately, most developers
> choose not to use those tools.  Instead, they choose
> Borland C++, Microsoft Visual C++, etc., because these
> have the most advanced classes and widely supported third-party
> libraries in the world.
> 
> 	< 	Mac DO NOT SCALE.  Only Unix scales.  So, 
> 	here's your clue to move to Unix!>>
> 
> Ah, without being too harsh, no one is calling for people
> to server on DOS/Windows or the Mac -- that would be
> foolish for large-scale systems.  I'm talking about
> Windows NT, which is quickly becomming the worlds
> most popular system for client/server computing.  The
> entire CompuServe infrastructure (and soon the Microsoft
> Network) is a distributed network of NT servers running
> SQL Server and various object-database tools such as Poet.
> 
> NT is a fully multi-threaded multi-tasking 32-bit OS with
> all your favorite UNIX tools.  It is also the only 
> server-OS (outside of Tandem and SCO 5.0) that truly
> supports Semetrical Multi-Processing (SMP).  You can scale
> up to dozens of processors accross a distributed network.
> In fact, Tandem, long the leader in distributed transaction
> processing systems -- is slowly abonding their own UNIX
> flavor in favor of NT.  They're reading the market correctly,
> I think.
> 
> You can pick up a copy of NT at your local software store for
> $99.
> 
> 	< 	Web.  The Web is only one interface to a database.
> 	Thus, I dont think we need to leverage database
> 	technology -- we need to leverage all those programmers
> 	you gave us the Web!  (By the way, these are not
> 	Dos/Windows programmers -- they are Unix programmers).>>
> 
> No, the point is that what has driven the commercial desktop computing
> world has been first, the desktop database revolution manifested
> early on in DOS database tools, and now in more sophisticated client/server
> tools.  This created an enormous pool of competent software developers.
> They are all ready and available for developing relational-database
> apps on the Web.  Second, there is a generation of visual programmers
> in the Windows world (100,000s) that are eager to use their talents
> on the InfoBahn.
> 
> The base of UNIX programmers that have been active on the Internet
> is tiny in comparison with the market of software developers in the 
> Windows world.  It would seem to me both short-sighted and myopic
> to not tap this enormous base of talent and knowledge.
> 
> 	< 	Pentium PC, Jeremy.  Also, can you blame Sun
> 	for developing on it's OS first?  There is 
> 	a reason James Gosling is working at Sun and
> 	not at Microsoft, Jeremy - freedom to create
> 	new things.>>
> 
> There is no comparison.  An open market in hardware
> in the PC world has created substantially cheaper
> alternatives.  Comparable SUN systems cost $5000 to
> $20,000 more.
> 
> Dual-processing 120Mhz PCI/SCSI systems run under
> $5000.
> 
> 	< 	the machines that "serve" the world - they
> 	are the client machines.  Of the other 5%,
> 	a "good number" run Unix.  We can start a 
> 	real revolution if we can get the other
> 	95% to run Linux!>>
> 
> Yeah, that pretty much seems like what the corporate
> and consumer markets are begging for.  Maybe they're
> just blind and need to be led . . . 
> 
> > (though, if folks read last weeks Info-World --
> > http://www.infoworld.com -- they might learn some
> > interesting and frightening things about Java's
> > history).
> 
> 	<< 	find "frightening"?>>>
> 
> In essence, the story read that Java, along with
> about 10 other projects internal to Sun, were just
> about to get the Axe when the Java team decided
> to independently release their code and unleash
> a Web about the system.  Having to save face,
> Sun turned an embarassing internal mutiny into
> a PR success by pushing Java as a nex-gen distributed
> network application.  The story continues that
> the technology is no where as far along as any
> hopes, at this point.
> 
> 	< 	new exciting Web applications is NOT the
> 	big expensive hardware.  The REALLY wonderful
> 	stuff will come out of new young companies 
> 	and from independent individuals -- ALL of
> 	whom can afford Linux machine with development
> 	tools.  Its under $6k.  Is that cheap enough?>>
> 
> $6k plus the cost of learning, support, etc.  Companies
> just are not going to deploy mission-critical systems
> on a totally unsupported OS and on totally unsupported
> software.  They have enough of a nightmare trying to
> train and support very well designed and documented
> software and systems.
> 
> 	< 	"Windows talent" doing Web applications.
> 	Windows is "corporate", and corporate is not
> 	the way to create exciting new apps.>>
> 
> AH, on this I think you are sadly mistaken . . . time
> will tell, eh?
> 
> Best,
> 
> Jeremy Allaire
> 
> 

	Rex Ballard
	Standard & Poor's/McGraw-Hill
	Opinions expressed do not necessarily reflect
	the Management of the McGraw-Hill Companies.







From rballard@cnj.digex.net Thu Jul  6 09:03:16 1995
Status: O
X-Status: