Conference 7.286::digital

    No....we've also got Intel systems, which ship in much higher volumes
    than Alphas from Digital ever will.
    
    IMHO
    Mark
    

RE: 3512.0 by GVPROD::DOIGTE::Chisholm

> This weeks Business Week says that Intel has 80% of the w/w microprocessor
> market and that Motorola, IBM, HP, SUN, MIPS and DEC, plus a few others in
> Japan are  fighting for the remaining 20%.

Intel has this market share based on the x86 family.  As long as the
software written for the 186,  the 286,  the 386 and the 486 will run on
the next generation  (Pentium is the "586",  but is not called that for
legal reasons),  users have a real motivation to stay with Intel.  The 
problem for Intel is that the x86 family is at near end of life.  For a
list of instruction set issues,  a x86 will run at 1/4 the speed of an
alpha,  all other things equal (number of transistors,  etc).   Intel can't 
change the instruction set without making the old software non-functional.  
Intel can't continue to just push x86,  as users are switching to new
microprocessors.  Intel knows this and has killed the P7 (the "786") as not 
competitive.  

Intel needs to do one of several things:

1) Design a new microprocessor with an instruction set something Alpha like,  
or "RISC",  and play catch up with Alpha,  MIPS and PowerPC.  Playing catch up 
isn't fun,  at best,  and about the time Intel might have catch up,  the 
generation after "RISC" will be starting up.  What then?

2) Try to define the next generation processor.  While a "RISC" machine has
a roughly four times advantage in speed over a "CISC",  a "Very Long
Instruction Word" machine or perhaps some other possible machine will
probably have a similar advantage over "RISC" machines.  The problem here 
is you spend a lot of dollars and might not get a marketable machine.  DEC
did this with uPrism,  and did get a very marketable machine in 1988,  which 
was killed to aid marketing the VAX9000.  

3) Regardless of what they pick above,  keep making and selling x86 chips.
Just as digital is still making and selling VAX and PDP-11 products.  


> Correct me if I'm wrong but are we betting our company on winning this
> battle ?

The Alpha is going to get a significant market share of the "RISC"
microprocessors,  which is going to be a majority of the market in 5 years
or less.  This means that the bet will probably pay off in terms of the
survival of digital.  If we don't bet on alpha,  digital's survival is much
more in doubt.


Phil

    We also generated larger revenue from Intel-based sales for the past 2
    quarters; as I recall it was Intel #1, Alpha #2, and VAX #3 in both
    quarters.
    
    JW

    	Intel are the dinosaurs with their CISC systems, and they haven't
    got an alternative. In 5 to 10 years one of the current RISC systems
    will have won out, and I have shares to show that I believe it will be
    AXP.

    Could someone comment on the HP/Intel alliance and the "very long word"
    (I've forgotten what it's actually called but VLIW seems to stick in
    my mind; it's in the same Business Week article referenced in the base
    note) technology and how that might play out vis-a-vis Alpha, PowerPC, 
    and the other possible players.
    
    HP/Intel think they've got the winner with VLIW (hope that's the right 
    acronym) technology and that they'll be able to retain the majority 
    of that 80% share Intel has now.
    
    -j

HP and Intel haven't said what technology they'll be using - there's
speculation that they intend to use a VLIW (Very Long Instruction Word)
architecture, mainly because HP acquired some people and technology from
a defunct VLIW company called Multiflow.

Somewhere (ALPHANOTES?) I saw a detailed analysis of the announcement, which
is an accomplishment considering that the announcement itself was just
smoke.  How anyone can take this alliance, with its seemingly impossible
to meet goals, seriously is beyond me.

						Steve

re Note 3512.2 by BOXORN::HAYS:

> 1) Design a new microprocessor with an instruction set something Alpha like,  
> or "RISC",  and play catch up with Alpha,  MIPS and PowerPC.  Playing catch up 
> isn't fun,  at best,  and about the time Intel might have catch up,  the 
> generation after "RISC" will be starting up.  What then?

        In what sense would Intel have to "catch up"?

        Do you mean catch up with the technology?  Recent history
        demonstrates that the second or third comer to a technology
        isn't at that much of a disadvantage, as long as they are
        aggressive and invest heavily.  They even have the
        opportunity to avoid others' mistakes and take advantage of
        others' market-building.  Besides, Intel is not a
        semiconductor technology lightweight; they have invested in
        RISC for many years.

        Do you mean catch up in the market?  If Intel can in any way
        leverage their current market presence (and I'm sure they
        will try), it might not take them long to surpass whoever is
        the leader of the RISC 20% of the market.

        Of course they have to catch up.  But sometimes an elephant
        can dance, e.g., IBM had to catch up in the PC market in the
        early '80s.

        Bob

RE: 512.5 by WRKSYS::LORD "Our forgetteries are in fine working order." 

> Could someone comment on the HP/Intel alliance and the "very long word"
> (I've forgotten what it's actually called but VLIW seems to stick in
> my mind; it's in the same Business Week article referenced in the base
> note) technology and how that might play out vis-a-vis Alpha, PowerPC,
> and the other possible players.

VLIW is a next generation (ie:  after Alpha,  MIPS and PowerPC) machine.  
"VLIW" stands for "Very Long Instruction Word",  and it deals with a
problem in current machines with instruction pipelining clashing with
multiple issue of instructions.  While it should allow for higher
performance,  it's not clear if it is the whole answer for the next
generation of processors.  


Phil

    FWIT -
    
    Please see the last two issues of Byte Magazine (Oct'94 & Nov'94) for
    discussions on RISC processor announcements.  The October issue covered
    our 21164 chip, and the November issue reviews the major competition:
    o  From Workstation Vendors (64 bit competition)
       -  MIPS / T5
       -  SUN  /  UltraSPARC
       -  IBM-Moto / PowerPC 620
    o  From INTEL "wanna-bees" (Pentium competition)
       -  AMD  /  K5
    In the November issue is also a discussion on the VLIW issue
    (pp 287 - 288).
    
    My conclusions are that (a) everyone else in the workstation vendor
    group *except* HP will have volume production of 1st-gen 64-bit 
    processors by 2HCY95, or 1HCY96, and that (b) Intel may need VLIW
    in order to perform x86 emulation fast enough to keep up with the 
    competition (the AMD-K5 discussion is of particular relevance here, 
    because it looks like they have done something similar in their x86
    to RISC translation process).  Given all this, I still expect HP
    and INTEL to make some sort of announcement for the 1996 time frame,
    even if it turns out to be some form of "marketecture".
    
    Charles              
    
     

Re: .8

I'm sure Intel and HP would like you (and everyone) to think that VLIW is
"next generation", but in fact it is very old technology that just didn't
pan out.  Nobody I know of has been successful marketing a VLIW system -
they couldn't compete with RISC or even good CISC implementations.

					Steve

    re: HP/INTEL alliance
    
    One other thing that nobody here has mentioned (and more FUD for the
    fire)...
    
    HP is claiming that this "POST PA-RISC" chip (the HP/INTEL chip) will
    be BINARY-COMPATIBLE with both its current PA-RISC architecture and
    with Intel's x86 architecture.
    
    In the words of Bob Palmer, "That's nonsense."
    
    It seems that HP (and Intel?) can say anything they like right now and
    the press will just believe them.  We should be challenging this
    statement PUBLICLY (in the press) every chance we get.

    Well, I think that it is possible for them to maintain binary
    compatibility.  It'll make it chock full of gates and painfully
    slow.  Albeit, that kind of defeats the purpose ...  An image of a
    smoking golf ball comes to mind ...
    
    Steve

    IDC says that in 5 years 90% of all commercial UNIX applications will
    be running on the 4 major RISC platforms (ALPHA, PA-RISC, SPARC,
    POWER-PC).  Where will this leave Intel in the commercial UNIX market?
    
    They don't rule the world, just part of it.
    

    Why doesn't the 'obvious' solution of two CPUs work?
    
    It doesn't have to be a dual-processor system, it could be
    a tag-team arrangement like was done on the Mipsfair machines,
    which had a fairly powerful onboard CVAX just to get the system
    running (and to run diagnostics).  Only one CPU runs at once
    and when it detects it's got code for the 'other' one, it flips
    a toggle which freezes it and starts the other one.
    
    (Easier when you don't have pre-emptive multitasking!).
    
    If you want to use the right words you could call it
    a 'compatibility coprocessor'.
    
    Is this a bad solution because (a) the component cost is
    thereby increased in an extremely cost-sensitive market,
    or (b) because it requires Microsquish to write some clever
    software?
                     
    

RE: .13

Intel may only rule part of the market, but it's the highest-volume 
part of the market.  By comparison, the entire Unix market is down in 
the statistical noise.


RE: .14

Co-processor solutions (1) increase box cost in an extremely 
price-sensitive market with low margins, and (2) would require 
Microsquish software support, and Microsquish isn't exactly tripping 
over its own feet in the haste to provide it.  Microsoft's answer to 
moving to RISC is Windows NT, and Microsoft doesn't really care which 
of the hardware vendors wins out in the end.

--PSW

    re - HP and intel can say anything and get away with it....
    
    Remember, we said that the VAX 8800 was a 'supercomputer' (I still have 
    some old literature) and expected the market to beleive it -
    unfortunately they did, and were very disappointed when the performance
    claims did not pan out....
    
    The beginning of the 'disbeleif' of DEC claims. (Jupiter
    notwithstanding...)

    
    	Microsquish (I really like that term, much better than Cutler's
    Microsloppy) owns PCs thru Windows. Usquish plans on owning the RISC
    market with WindowsNT at the server. Are these guys geniuses, or what?
    	I suggest we tune Alpha to run NT like two bats out of hell. Our
    current focus on recruiting NT ports is probably a very good move, and
    we should do more.
    	Unix as statistical noise is only partially correct. Measured in
    units, yes. Measured in dollars, and especially margin dollars from a
    hardware standpoint, no.
    	Now a Digital/Usquish UNIX/NT "Universal Server" could be a real
    show stopper.
    
    		the Greyhawk

    Dunno who invented "Microsquish", but *I* stole it 
    from Berke Breathed - the strip where Gates's mind
    was projected into Bill the Cat.

    re.13 Commercial UNIX
    
    Intel has been winning the Commercial UNIX VOLUME business with SCO.
    However, from a revenue $$ perspective, they're way down the list.
    (HP has over 40% of the Commercial UNIX market using revenue as the
    criteria).  With the rapid growth of HP, IBM, SUN, ATT/GIS and perhaps
    even Digital, eventually, in the Commercial UNIX market from a volume
    perspective, the Intel lead is eroding (maybe SOLARIS on Intel will
    keep things moving for a period of time).
    
    It remains to be seen how the Intel/HP next generation RISC technology
    and platforms will do in the market, and that will determine Intel's
    Commercial UNIX future.
    
    (We had a chance to do much better in this market a couple of years
    ago when Dom LaCava created the Commercial UNIX Marketing Group to
    attack this fast-growing segment of the UNIX market.  Unfortunately,
    soon after, the Central Engineering Managers decided that was not
    a market/product focus and the Group was eliminated....we could've
    been a contender 8^)). 
    IMHO
    
    Mark Silverberg
    
    

RE: 3512.10 by QUARK::LIONEL "Free advice is worth every cent"

> Nobody I know of has been successful marketing a VLIW system - they 
> couldn't compete with RISC or even good CISC implementations.

RISC machines are faster than CISC because prefetching of data is done as
soon as possible under the control of the software (and other reasons as
well).  A CISC machines can and do prefetch data under hardware control,  
but such hardware prefetch is quite complex,  meaning slow,  hard to design 
and test,  and takes lots of resources that could be used in other ways to 
speed up the machine.

To speed up a RISC machine,  several instructions can be "issued" (started)
at the same time under hardware control.  Hardware multi-issue (beyond a 
small number) requires very complex hardware,  meaning slow,  hard to design 
and test,  and takes lots of resources that could be used in other ways to 
speed up the machine.

VLIW is ONE way of making a software controlled multi-issue machine.  There
are others,  and Intel/hp are taking the risk that one of them might be
much better than VLIW.  While I think Intel/hp is wrong,  digital isn't
doing a next generation design that I have heard of.

Now,  VLIW machines COULD BE MADE TO BE 100% BIT LEVEL COMPATIBLE WITH
SEVERAL EXISTING MACHINES.  How?  Simple.  Let's design a simple VLIW
machine that is 100% 8085,  Z80,  6502,  PDP-8 and 6809 compatible.  These 
are old 8 bit microprocessors,  for ye youngsters,  other than the PDP-8, 
which is a 12-bit minicomputer.

Instruction word formats:
|<8 bits>|<8 bits>|<8 bits>|<-8 bits->|
|<-8085->|<-6502->|<-6809->|<control >|

|<8 bits>|<--12 bits-->|<--12 bits--->|
|<-Z-80->|<---PDP-8--->|<--control--->|


The hardware is one of each microprocessor,  plus some address and control 
logic.

To run an 8085 program,  the existing code would be put into the left
column,  and the control field would be set to enable the 8085 machine, 
and to give the 8085 machine control over instruction address space (on a
cycle by cycle basis).  As some 8085 instructions require data cycles,   
instructions to other machines could be interleaved into existing programs.
To run a 6502 program,  the existing code would be put into the second
column,  and the control field would be set to enable the 6502 machine,
and to give the 6502 machine control over instruction address space.  This 
example is to show 100% bit level compatibility only,  and would have only 
a very limited performance advantage over a single issue machine.  


Phil

VLIW processors are very complex by nature and to get anywhere close to
peak performance out of them, hand-tuned code is generally needed.  Compiler
technology is having a hard enough time getting good performance out of
RISC - VLIW is several times more difficult than that.  Also, few applications
can take advantage of the type of parallelism that VLIW offers.  RISC offers
most of the advantages of VLIW without many of the problems.

As I mentioned, VLIW has been done before.  Its promoters promised great
performance which never materialized, and this was WITHOUT any requirement
for compatibility with two completely different existing architectures.

We designed Alpha to take us for at least 25 years, and I think it has every
promise of doing so.  Don't let vapors cloud your mind.

					Steve

    	Almost all CISC machines were VLIW if looked at from the microcode
    level. On the PDP-11/60 the instruction length was more than 50 bits
    (don't remember exactly - ask Rich Lary) and the VAX/11-780 was 98
    bits. Those are the only two systems I can remember off-hand where we
    offered VLIW programming to customers, but we have 20 years of
    experience with it.

    The IBM 360/50 had a microinstruction width of 80 bits I think.

    VLIW machines are from 256-1024 bits.  i.e Multiflow has 256.  They usually
    have very high cycles per instruction and therefore need wonderful
    compilers with brilliant branch prediction.  But if that works they get
    pretty good performance.  If a bunch of 32-bit instructions get
    compacted into a VLIW the apparent cycles per instruction get to be as
    good as .33 in Multiflow, which is certainly in the ALPHA ball park.  
    But they need
    the great branch prediction.  And if they don;t get the code density
    (compaction) from compilation, they run a lot of no-ops in the VLI.  
    
    They don;t have the problems of run-time syncronization or resource
    scheduling that superscalar machines have, but again that means all
    of that has to be done by the compiler before run time.
    
    Read J. A. Fishers 1983 paper in the Proceedings of the 10th? SYmposium
    on computer architecture.
    
    I think it is one  of those great ideas whose time may not come for a
    while.
    	
    Alpha is a really decent machine.  WHen we get OSF and NT and our tools
    tuned up and when people start to take advantage of 64 bits, we should
    really shine.
    
    

    In today's WSJ, Motorola announced that they would provide W/NT on the
    Power-PC chip.  This was considered a boost to Microsoft and a jab at
    Intel.  The article went on to explain the W/NT market and Microsoft's
    plans for Windows and W/NT.  In said that in the future, when a
    customer can run their legacy Windows applications on W/NT, then they
    will not be bound to Intel and will most likely pick the system that
    provides the best price/performance (RISC).  It also said that the RISC
    chip chosen will have to more than 1.4 times faster than a Pentium. 
    This was an obvious swipe at the 133 Mhz Power-PC chip vs. the 90 Mhz
    Pentium.  The point of the article was that as W/NT is ported to other
    chips, becomes more successful, allows you to run Windows applications,
    runs faster on RISC chips, then Intels hold on the market is
    threatened.
    
    300 Mhz Alpha is 3.3 times faster than 90 Mhz Pentium.  A recent Alpha
    customer survey conducted by Digital found that the number one reason
    for buying Alpha was cost of ownership (price/performance) and that is
    how Digital should market Alpha.  
    
    This is just the first inning folks

    RE: .26
    
    Wrong.  The #1 reason to buy any computer is to create value for your
    firm by solving a business problem or making yourself more competitive
    in the marketplace through a lower cost structure, higher revenue
    generation, higher margins, or lower cost of capital.  It has
    little to do with the speed of silicon, but rather how it is applied in
    concert with a SOFTWARE APPLICATION TO SOLVE *THE* BUSINESS PROBLEM at
    hand or creating shareholder wealth.
    
    It's a commodity folks.  The technology is being sold at rock bottom
    prices with margins so low you wonder how anyone can make any kind of a
    profit even in the best of times.  If you have a solution to the
    customers problem that supports goal #1 better than anyone else in the
    marketplace you have just created a cash engine that will chug out
    money like you can't believe.
    
    		Maverick

    RE: .27
    
    Hear! Hear!

re. 26

RIGHT!
 
This is an excerpt from the report pertaining to the reason customers buy 
Alpha.  It was produced based on the responses of BUYING CUSTOMERS, not the 
repsonses of Digital's employees.

Subj:	MAJOR FINDINGS AND IMPLICATIONS: ALPHA CUSTOMER STUDY                  1

      Survey recently conducted for us by the Boston Research Group.


         Executive Summary of 1994 Alpha Customer Feedback Study
                   with implications for the business

------------------------------------------------------------------------


                   -- The Reasons for Buying Alpha --

FINDING:       "Performance", "price", "cost of ownership" and 
               "service/support" dramatically dominate the reasons for 
               buying Alpha in a competitive environment, far outweighing 
               "choice of operating system" and "scalability".

IMPLICATIONS:  Any loss of leadership, or loss of awareness of 
               leadership, on these attributes will strongly affect 
               Alpha's momentum.

	       In low-end server and workstation market segments where 
               Alpha may not have price leadership and where CPU 
               performance leadership matters less, Alpha won't win often 
               enough to establish a defensible position.

	       We should consider raising the visibility on "cost of 
               ownership" messages in our Alpha advertising and 
               communications. 
-------------------------------------------------------------------------

    
    	My continual amazement is in why we insist (that is to say
    Digital's management structure) that what customers say has no validity
    as to what we do. Even my three year old has got that part down. If
    Mommy or Daddy is focused on this is the way *we* want the job
    accomplished, she performs accordingly. Just substitute M&D for our
    customers, and voila. Am I missing something here, or did we just get
    stupid along with arrogant?
    
    		the Greyhawk

    Arrogance is just another form of stupidity!  So you answered your
    own question. 
    
    In fact, arrogance is the end result of GETTIG AWAY with stupidity!
    
    p/t
    

RE: 3512.22 by QUARK::LIONEL "Free advice is worth every cent"

> We designed Alpha to take us for at least 25 years, and I think it has
> every promise of doing so.  

You can still buy PDP-8's,  (not from digital).  Digital still sells PDP-11's
as well.  Alpha is current technology today,  but it's not likely to be
current technology in 10 years.  Why?  Simple.  Technology changes.

A simple fact of more and faster transistors per chip.  You just can not take 
any design and scale it up.  There are two competing ways to use the more
transistors that can be put on a chip.  You can make the chip run faster or
you can put more things on a chip.  In both cases Alpha loses.

In the first case,  hardware multi-issue costs are of the order of k1*n^2,  
and software multi-issue costs are of the order of K2*n.  As n rises,  there 
will be a point in time when software multi-issue machines will win.  My
guess is five to ten years from now.

In the second case,  an Alpha is not a good fit for a System On a Chip. 
Put (say) 4 M to 16 M bytes of RAM and a simple processor on a single chip.
Fast,  as main memory access times would be greatly reduced.  Very Cheap 
system.  Five or less years away.


> Don't let vapors cloud your mind.

Don't fall in love with a cart when everyone starts to fly.


Phil

re: .26-.29

I think that saying price/performance is the number 1 reason for buying assumes,
given that we're talking UNIX here, that the application the/a customer has 
identified to solve their business problem runs on our and competing platforms.

Bruce

>                      Even my three year old has got that part down. If
>    Mommy or Daddy is focused on this is the way *we* want the job
>    accomplished, she performs accordingly. Just substitute M&D for our
>    customers, and voila. Am I missing something here, or did we just get
>    stupid along with arrogant?
>    
>    		the Greyhawk

     Isn't it interesting that the only place you appear in your
own example is on the telling (as opposed to listening) side?

    re. 33
    
    I would say that you have it just about right.  Some existing VMS
    customers may choose to add new applications on VMS, some on OSF/1.

    
    	.34
    - Nice flame; however I suggest you re-read the note. No telling is
    involved, just accomplishing agreed tasks according to plan. Just
    like what our customers expect of us. Last I remember we live in an
    adult world where expectations equal performance.
    
    		the Greyhawk

    
    	Also let me add a little more....
    
    	Today I received my annual report from Solectron, a $1.4-billion
    contract manufacturer of PWBs. A quick quote -
    
    	"*Structured for customer success.* Solectron's corporate structure
    provides centralized purchasing power, global account management,
    regional operations management, and local customer support. Our goal is
    to use this infrastructure to 'leverage globally and excute locally',
    giving our customers the best of all worlds."
    
    	This company also has monthly, and in some cases-weekly, customer
    satisfaction surveys on quality, delivery, communications, and service.
    Any grade less than a B requires a specific Solectron action plan by
    all levels within the company associated with that customer.
    
    	BTW, their SGA is 3.6% of revenue, margins are only 10%, and profit
    is 5.8% of sales.
    
    	As Tom Peters' preaches constantly, "Have a passion for the
    customer, empower employees, and the profits will follow."
    
    	And some noters have a problem with me? Come on, get real, do you
    actually know someone more passionate?
    
    			the Greyhawk

Hi folks,

the best entry in this whole discussion was 3512.27. The processors 
don't matter - it's the business that counts. A CEO (I think it was 
Apple) once said "it doesn't matter what's powering it - it could be a 
hamster wheel". I liked that phrase.
If you look at most companies in Germany we have basically three types:
i)   Innovative, state-of-the-art computing ie. Client/Server e.g. Audi, 
     BMW, VW to name a few
ii)  "Wait-and-see" companies still using main-frames and planning a 
     move to newer technology "after it's been proven". We're talking 
     about the majority here because they'd rather be safe than sorry 
     e.g. most major banks, insurances, chemical companies
iii) "As long as it does the job", "never touch a running system" companies
      e.g. Siemens, Avon cosmetics, Maizena (CPC), electricity suppliers 
      the list is endless
BUT: there is no correlation between the technology a company is using 
     and its financial performance. Or in other words: you can have a (i) 
     doing badly and a (iii) doing well.

Mitch

    
    
    
    
    
    
    
                      -- The Reasons for Buying Alpha --
    
    
    reference 3512.29
    
    
    FINDING:       "Performance", "price", "cost of ownership" and 
                   "service/support" dramatically dominate the reasons for 
                   buying Alpha in a competitive environment, far
   		   outweighing 
                   "choice of operating system" and "scalability".
    
    IMPLICATIONS:  Any loss of leadership, or loss of awareness of 
                   leadership, on these attributes will strongly affect 
                   Alpha's momentum.
    
    ========================================================================
    
    Question:
    
    Since more people have not bought an Alpha than have bought one,
    what is the message for the folks who have been walking away?
    
    Why do we lose when we lose, despite all of these virtues?
    

    re: .39
    
    Exactly.  I want to see -- The Reasons for Not Buying Alpha --.  Then
    we can focus on eliminating those reasons.
    
    	Steve

    
    	The reasons are all over these files people.
    
    	1) Poor, or non-existent, technical support close to the customer.
    
    	2) Digital cannot deliver product on time; it cannot bill the
           customer correctly for their purchase; it cannot maintain the
    	   purchased items without causing a hassle for the customer;
    	   and it constantly "grates" customers by its on-going "internal"
    	   wars about who gets credit for the sale.
    
    	3) Product testing of software is marginal. More "things" don't
    	   work than do - as far as the customer is concerned. This mainly
    	   occurs because Digital people are stretched too thin to properly
    	   support customer's operational questions within the CUSTOMER'S
    	   timeframe.
    
    	4) Digital salespeople, and our distributors salespeople, are
    	   poorly trained, badly managed, focused on extremely short-term
    	   goals (has anyone anywhere ever heard of a sales compensation
    	   plan measured in *QUARTERS*?), and in fear daily of being
    	   TFSO'd.
    
    
    	I'm not the least bit surprised customers' aren't buying Alphas -
    	today, sportsfans, they actually have choices. So we better stop
    	being smug and arrogant and get our butts in gear. The customer is
    	always right.
    
    
    			the Greyhawk

    re -.1
    
     and how to go about fixing them. ics just cannot fix these problems
    by themselves. Even groups of us are not going to fix these problems
    you describe. 
     
    <<<	1) Poor, or non-existent, technical support close to the customer.
    
    ics and CSCs only support a described set of products. Directions to
    come from corporate on product production. See some evidence on this
    now with the software strategy. IMHO we must move faster.
    
    <<<	2) Digital cannot deliver product on time; it cannot bill the
    <<<    customer correctly for their purchase; it cannot maintain the
    <<<	   purchased items without causing a hassle for the customer;
    <<<	   and it constantly "grates" customers by its on-going "internal"
    <<<	   wars about who gets credit for the sale.
    
    the ones that make the sale get the credit. end of story.
    revamp billing. Send bill, get check or cash.
    deliver product on time. step one, do I have a product ready to sell?
    step two, is it in quantity? yes no, on sale/off sale.
    
    <<<	3) Product testing of software is marginal. More "things" don't
    <<<	   work than do - as far as the customer is concerned. This mainly
    <<<	   occurs because Digital people are stretched too thin to properly
    <<<	   support customer's operational questions within the CUSTOMER'S
    <<<	   timeframe.
    
    software is a key component of our survival, the sooner someone
    realizes boxes don't move without it, the better off we will be.
    we need to become a highly focused software company. see gripe 1.
    
    <<<	4) Digital salespeople, and our distributors salespeople, are
    <<<	   poorly trained, badly managed, focused on extremely short-term
    <<<	   goals (has anyone anywhere ever heard of a sales compensation
    <<<	   plan measured in *QUARTERS*?), and in fear daily of being
    <<<	   TFSO'd.
    
    we have sales people? The ones I know of are excellent but stretched
    way to thin. get longer goals. say "no" to the compensation plan, find
    a consensus among your peers. demand better. educated our distributers.
    
    Indeed a daunting task, but fixable. Present this to senior managers,
    find out who ownes these problems, hold them accountable. We have the
    products to compete! Just the internal mis-directions and strife remove
    our focus. I have a problem with a customer right now that is trying to
    send digital a check. The poor gentleman cannot find anyone to take his
    credit card number. Just plain incredible. Start with picking up the
    phone and seeing what works these days is interesting.
    
    -Mike Z.
     on-site at customers playhouse since 4am on a saturday.....
     customer satisfaction in process '^)