Author: Vincent Diepeveen
Date: 10:57:03 10/24/03
Go up one level in this thread
On October 24, 2003 at 11:22:33, Daniel Clausen wrote:
>On October 24, 2003 at 10:35:43, Vincent Diepeveen wrote:
>
>>On October 24, 2003 at 04:50:55, Daniel Clausen wrote:
>>
>>>On October 23, 2003 at 19:42:59, Vincent Diepeveen wrote:
>>>
>>>>On October 23, 2003 at 05:55:12, Daniel Clausen wrote:
>>>>
>>>>i've tested so many toolkits and environments and really
>>>>very *little* even *works* both in linux and windows.
>>>
>>>Agreed.
>>>
>>>
>>>>then possible sales under linux.
>>>>Yes linux is great, but selling something under it????
>>>>
>>>>It's simply NAIVE to guess that porting an application to *nix will sell some.
>>>
>>>The chicken and egg problem again. A good start though is for example to provide
>>>a Linux version together with the Windows version, as DeepSjeng and hopefully
>>>Ruffian and Tiger will do in the future. This way you slowly make Linux a viable
>>>option for the computerchess-interested community. If it is big enough, you may
>>>even try to sell something in the future. It's a long way, and it requires
>>>companies/people who think a bit and don't just look at the number of sells by
>>>tomorrow evening.
>>
>>That's all console shit you talk about.
>>
>>That's like saying linux is a succes because there is available a left hand
>>hockey stick for people with 4 fingers and no thumb.
>
>Did you read what I was writing? I'm trying to show a way how companies can
>slowly attract a big enough customer base on Linux, which would be willing to
>pay for a product.
>
>I realize that a company that sells a Linux product (not just Linux-version as
>an unsupported add-on in the Windows-package) needs a certain customer base in
>order to
let's distinguish 3 important sectors
a) custom made software (for 1 client), where you build it such that
it is cheaper for your client to use and where you can take into
account what your client wants to install or has already installed.
b) expensive restricted mass market software (where the
developer can restrict customers in what hardware / software
to install because their software is that important
for this company; for example certain databases and administrative
and product software is doing this)
c) mass market software, where the client is king and where everything runs.
The latter category is what we are talking about. The earnings at mass software
are real little when compared to the first 2 categories.
That means in short that if 20% of your customers is going to ask questions or
complain about your mass market software to your helpdesk, that you are bankrupt
simply.
At the moment if someone would release something for linux i assume that
currently helpdesk is going to catch many questions from many customers and for
sure more than 50%.
So if earnings are $10-20 a product, then everyone realistically will be capable
of predicting that your helpdesk can't handle that.
Most of the questions will of course be related to simplistic OS questions.
However helpdesk personnel in general is not sitting there because they are so
very good. They sit there because they are so very cheap to answer the
questions.
So where i could perhaps answer 60 questions an hour, hands down, the average
helpdesk 'office manager' as you know in such matters can handle 6 questions an
hour, especially because each question requires several emails usually.
Now linux because it's so hard to use in console mode for the average user, is
of course completely impossible to be using a helpdesk for. There is even
difference between capitals and small letters, which is a real pain in the ass
knowing that many files are a mixture of capitals with small letters.
No way to explain that to the average user easily without taking quite some
time.
That's major problem 1 for linux.
Imagine that Cardoza chess or chessmaster would be released for linux native.
And let's *assume* now, that they would sell the same amount of copies they sell
under windows.
So for cardoza chess that would be around 500000 copies.
Imagine that you get 450000 people asking beginner questions and 50000 users
asking very advanced questions (system admins very used to linux are known for
raising difficult subjects).
>(1) Hire people who build a Linux software-package which is tested on the major
>distros (like RH9, Suse9.2, Mandrake_whatnot)
The vaste majority of linux distributions used are older distributions.
For example i'm using SUSE 8 here (to install i had to download *gigabytes* and
i installed it from harddisk without creating cdroms) and by doing that i
already use a fairly new distribution. Of course i also installed gcc 3.3.2
I know several using redhat 7 and redhat 8. A single person who shortly used
mandrake and right now some weird name i never heard of.
In germany most use Suse and the cheap dutch usually use Redhat because you can
download the ISO's freely.
If supporting win98, win SE, win 2000, win 2003, win XP is already a pain in
itself, supporting all linux distributions is quadratic more important.
Further if we say 'linux' then many also assume that under that is also
'freebsd', which is what most stand alone internet servers run in companies.
So before releasing a product you must not only test it for quadratic more
distributions and releases and product versions than windows, you also must port
it to freebsd and such systems to look professional.
So development+testing costs *are* higher than for windows.
>(2) Have people in the customer support answering questions [almost _all_
>products have a customer support]
>>>If you _sell_ a product for Linux, you surely have to make that the software
>>>installs/runs as flawlessly as under Windows. (*cough* :) That _can_ be done. A
>>
>>Nonsense, you have no *idea* what you are talking about.
>>
>>Have you ever tried compiling a linux executable from a graphical program in
>>*nix and tried running it at different linux distributions and kernels with
>>different *.so and different multithreaded libraries?
>
>I assume you're talking about "running the executable", not compiling it.
i asked 2 questions
a) whether you have made a graphical GUI under X.
I did once (apart from cross compiling diep's GUI) and i hope i never have to do
it again.
b) whether you have tried to run an already compiled executable under
different linux versions across europe.
This is a major problem.
>As for graphical programs running under different Linux distros, just download
>Mozilla - I'm pretty sure it works out of the box under all major distros. The
>GUI of Diep probably doesn't depend too much on the kernel either - I'm sure it
It does, because it has threads for the engine and communication protocols and
all kind of stuff i don't want to post here yet.
>work under my gentoo kernel as well. ;) As for different *.so - that's not any
>different than under Windows.
Forget gentoo, only the engine runs under gentoo, if you have kernel 2.4.x and
at least gcc 3.1 installed, so not intel c++.
>Please note that with "supporting Linux" I never mean "support every Linux
>installation there is on Earth". Of course a company can only test their stuff
>on major distros. gentoo for example would simply be impossible to test for as
>the system constantly changes. (on purpose)
What i meant to say is also this.
Imagine that the most popular 1000 applications/games/software used currently
under windows, that those companies all at the same time port it to linux now.
Now imagine that some linux user, illegally of course, goes install 150 of those
applications.
What chaos do you guess will happen?
Linux is not prepared yet to run hundreds of similar applications without the
person installing being a console hero.
Linux looks good at the moment because there is so little choice in software
currently when compared to windows toys and tools. The DLL system of windows is
a big failure trivially, but the 'everyone his own tool-with-non-existing
documentation-just-find-someone-who-knows-approach' mentality in the linux world
is even more sick from normal user viewpoint.
Already installing a few applications in linux requires 5 hours of online
consultation of documentations and bugfixes and especially grabbing a few linux
dudes at their hair asking how you must get it to work.
Distributions like gentoo require you to be online and having an unlimited and
big bandwidth internet connection.
Well to tell you the truth, i *never* succeeded in getting my Suse online.
If i would have managed that, then i would perhaps be toying in linux more than
i do now.
My audigy2 soundcard, no idea how to get it to work. it doesn't work by default
in the one biggest linux distributor on the planet, if not biggest: Suse.
My network card, they only sold 20 million or something from that card in europe
for $10, doesn't work by default. Becuase i'm a freak i know it's having a SIS
chipset. That took me 2 days to find out. So i know the kernel CAN support it
when i turn it on.
Because of all those problems on average linux takes 15 minutes to boot.
The average user is simply not swalling: "then get hardware that does work under
linux".
The user will reply: "grow up, this OS will never get something".
No one in whole linux community is responsible to create a driver for an audigy2
soundcard. No doubts that by now there will be some solution, but i'm not
prepared to browse for 1 day online to find a solution.
Now suppose that i may explain to users for my software why they cannot run a
linux version of diep with sound, and after a long period of spending time to
that user you find out the problem is that by default his great soundcard is not
supported in linux for whatever idiot reason. I'm not yet prepared to do that...
...and i would be really amazed if there is mass market producers who are.
>
>>I bet not. You are of course like all unix hackers used to simply compile it
>>with source code. Well that won't happen of course.
>>
>>No selfrespecting interface can do without multithreading unless you find stuff
>>like xboard good for 2003 standards.
>
>I can't really comment on the multithreading part here, but what's wrong with
>pthreads under Linux?
In older distributions, which still are the most used distributions, they are
kind of 'broken'. And i figured out that my testers online could not run new
diep engine executables compiled without having installed a new thread library.
Of course later i found out i can make the executable 2MB bigger and add that
functionality to the executable. Which is what i do now.
So the linux executable is a lot bigger than the windows executable is now.
>
>>I hope you realize what a real product team costs. If you run a team of a person
>>or 10 to produce a commercial product, then development costs get expressed in
>>hundreds of thousands, not in cups of volunteerly coffee.
>
>I do that very well. That's why I'm trying to show a potential _path_ to a
>bigger customer base under Linux. Customers don't grow on trees, the market has
What you ask is a company to sell for $50 return tickets to mars and you expect
someone to do just that?
>to develop slowly in this direction. And things like a Linux-version of Ruffian,
>CT, DS _help_ a little, at least in the computerchess area. Of course, other
>steps in other fields are required too...
that's only console executables.
no one except a few freaks in this forum care about console shit.
>
>[snip]
>
>
>>You have really *no* idea how hard it is to make a commercial product for a new
>>OS.
>
>Actually I do. Our applications run under Solaris, Linux and Windows.
So most likely not a mass market product.
>>All layout, all i/o, and so on. Additionally you must write your OWN libraries
>>for linux, because the ones available suck ass. Users in linux DO want to see
>>the same like their windows counterparts can see.
>
>It's hard to comment that, since you're so specific here. ;)
>
>
>>I was shortly enthusiastic about wxwindows and such stuff, but then i created a
>>bit more advanced application within a few days in that, and it suffered from
>>bugs everywhere as soon as i wanted stuff to auto-resize.
>
>For a small-sized to mid-sized product, I bet the cost to _support_ it is
>clearly higher than the development cost.
Please distinguish between mass market < $50 products and between expensive
products where you have a contract with each customer for support.
>
>>The same functionality i made in 1 day under windows and it worked bugfree.
>
>I won't argue that it's easier to build GUIs under Windows as opposed to
>Windows. There are people who can do it under Linux though too. Just look at the
Of course. From the many GUI programmers on the planet a good GUI programmer
let's say $100k a year. To port an application to linux takes around 1 year
and additional work.
So let's say $250k for the total of the port?
So if i would sell DIEP at a commercial basis graphical under linux,
i would want to sell at least 25000 copies.
So most likely what happens is that you sell 100 copies and then all those linux
freaks who are very well aware how to copy software, will all copy it
around.
So for every copy sold i get 100 questions in my mailbox, starting with Djordje
Vidanovic.
Are you realizing how much 25000 copies is for a mass market product under
linux?
It's impossible to sell that simply.
The reason is very simple. I'll give it you once and for all.
There is not a single shop on the whole world who wants to sell such a product.
Also not a single computerchess dealer is going to be so stupid to order more
than 1 copy of it.
So i would need to sell it online. Selling online 25000 copies? Just forget it.
>GNOME desktop and all the (GTK2-based) applications available now. So it's
>possible...I'm sure you spent much more time with the Windows tools than with
>the Linux tools, so I'm not surprised you're way faster in Windows.
How many man hours have been put into the GNOME desktop?
This is a crucial question because it must be answerred in order to calculate
something big to be released at a commercial basis for linux.
>>I know so many linux interested persons, and all of them share that
>>*everything* they ever do under linux, it *never* has to do with graphics.
>Maybe you know the wrong people? ;) Seriously, what exactly couldn't a person
>who installs RedHat or Suse on their PC _not_ do?
For example i can't edit my files in my favourite editor under linux.
You can ask at witnesses after my typing and browsing speed.
Which might explain something about why i can sometimes do a bit more than
others.
But i also browse at the same speed that i can type.
Scroll through my source code. And that in colors.
without using an emulator, forget any editor.
No i don't like vslick either. for unix standards it's perhaps great but it
sucked from my viewpoint.
I'm currently using multiedit 8.0
>>Sure they want to use cool graphics, but it is here where windows really has a
>>huge edge at linux.
>
>If that would be a killer, we all would use Macs. ;)
>
>[snip]
>
>Sargon
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.