The product, when introduced, was indeed clunky, but has quietly and steadily improved over the years. I'm a happy customer, having bought it twice for different customers.
I use a laptop most of the time, with a larger machine
at home serving as a fileserver and fallback. To keep
my mail and projects directories in sync I use unison,
reviewed here
You control the app, so you can easily log the response time of
the database, and measure it with a small but credible test load.
If you send the database a hundred or so requests at a rate that's deliberately
slow, you will get an average response time that's a good estimate of
the actual internal response time of the database. Let's say it's
1/10 second. With that number (alone) you can predict and plan for the performance you need.
On a uniprocessor, you will get a maximum of 10 transactions per second
before a queue starts to build and you start seeing delays due to queuing,
so for 1 through 10 TPS, the response time will be 1/10 second.
In fact, as you get closer and closer to 10 TPS, there is a higher and higher
probability that two transactions will come in at the same time, and the
queue will start to build before you hit 10...
After the queue builds up, every additional request that comes in
will need to wait before it get processed. T queue length is calculated using
Little's law, Q=XR, where X is offered load and R is response time. A
load of 50 requests would yield a queue length of 50* 1/10 = 5, and the average response time at that load would be (40*0.1 + 0.1) = 4.1 seconds
Voila! For a load of 50 TPS and a target time of 1/10 second, you need
a five-processor system, easily achieved with 3 dual-core AMDs.
Feel free to send me mail: I have a copy of Neil Gunther's "pdq" queue
solver and can easily compute what-ifs based on your measurements and needs.
See "Analyzing Computer System Performance with Perl::PDQ", at http://www.perfdynamics.com/iBook/ppa_new.html
Normal computer users aren't being served well by Windows, either. Dumbing
down an interface won't help my Grandma (actually a friend's grandma) at all:
she's a retired MD, businesswoman and druggist. Which is to say she's a normal person,
not a moron.
What she needs is a way to change something without writing a
program, something she doesn't know how to do. Therefor she needs
a graphical representation of the subject matter, such as her
patient records, and a way to manipulate and edit it symbolically.
As it happens, the company that did the research in that area is Xerox, and the
one which popularized is is Apple. Windows is a non-starter, so don't
make the mistake of copying it...
No, they just wanted something that could implement "time machine", their
backup-done-right proposal.
I expect Solaris to be switchd to GPL some time in the GPLv3 era, at
which point there won't be a problem porting ZFS to Linux. Not that
it was technically difficult to port it to Apple/BSD (;-))
As it happens, there are an amazing number of bad jobs
out there, and a moderate number of good ones. One often has
to tell your boss that you're conflicted but won't lave hime in the lurch.
Then you can start a job-hunt for a good job, and on finding one, negotiate a
reasonable notice period and hand-over.
Menu structures are common across different models
on
Office 2007 UI License
·
· Score: 2, Interesting
Many moons ago, I worked on a product which started out using
a "lotus 1-2-3" menu structure: one typed "/" then selected from a one-line
list of options by typing individual characters.
My Smarter Colleagues noticed that from the same data structure
we used for the lotus menus we could build PF-key menus, modern
cascading drop-down menus and right-mouse-button pop-up menus.
Which means that for any menu sequence of head->middle->middle*->tail,
you can change the visual appearance of the menu without changing the
application-level calls used to create it. And that in turn means you can make "ribbon menus" a user-specifiable "skin".
I used to work in a large international firm based in Germany, developing one of
their large-scale business offerings.
Some portions used open source software, notably for interoperability
with PCs, and we had to say so in the customer documentation.
Customers (and therefor our managers (;-)) wanted three things:
a CD they could hold in their hand
a company who would offer them a support contract, although they rarely would actually take out a contract
a printed, bound manual, or, preferably, an O'Reilly book.
If they had these three, the level of concern about the kind of software it was
fell very rapidly. If we only offered them two of the three, they tended to to be very dubious about the whole deal. You can imagine how that motivated us (;-))
As to support, the customers were very open to our company having a
relationship with the software developers, with us reporting bugs and proposing
patches. This in turn made our managers see the GPL as part of a quite normal business
relationship with the developers, and in some cases had us offering the
support for the open source components as part of our own service contract.
The Multicians had a set of rules that allowed
them to do "continuous maintenance" on production
systems without forcing everything to change at once.
The compatibility didn't last forever, but if you
were an application developer, you really only need to do
a QA run once a quarter to see if anything was in the
process of changing, and schedule a fix for the next quarter.
Your customers had a longer guarantee, probably a year or
so.
The idea was a lot like relational database theory:
you can always "add a column to the table" and you
could use NULLs to mark an old column as not containing
anything any more, although the actual technology
was major- and minor-version-numbers, and was
derived from some hardware-versioning research at MIT.
They still work: I used the same techniques in a Unix project, and
never had to have a flag day, although Edsel and I
were changeling the interface in question with wild
abandon (;-))
Actually, quite a number of folks who are my consulting
customers
use Darwin (really BSD) sources as the "reference copies" of
programs they're adapting for their own use.
This is in part because of the good quality of the code,
and the company which stands behind it. In part it is
because of the larger BSD community who stands
semi-invisibly behind Apple... some customers really
understand the strength of community. And finally, for
the license-paranoid, in part this is because of the
use of the very old and weak BSD license.... some customers
really
don't understand the community (;-))
Coming back to the main point of the discussion, adoption
of the GPL by well-known fortune-500 companies is a step
away from the world of Microsoft, SCO and FUD.
Definitely a change, and definitely for the better.
In a previous life, I had a smart-card for a badge, which
I shoved in a sunray x-terminal or a laptop as the"thing
I had", and typed a password as the "thing I knew", after
which I got my current session back.
If I needed to so somewhere else, I unplugged the card and
my session was saved. When I got there, I plugged back in again,
typed my password to the screen-saver and picked up exactly where
I left off.
I was very pleased with this scheme: it saved me hours of
frustration with AD kludgery and the string of crypto-keyfobs I
now have to cart around.
And it's GPL, same as Open Office (from many moons ago).
I suspect you'll see more of this: Sun uses specialized licenses only
when there is some legal reason to do so, where "legal reason"
can mean "Microsoft fork", "patent trap" or something less contentious such
as "we don't own that code, so we can't sublicense it".
... in enough english-speaking jurisdictions in North America that
library software companies arrange for their programs to only keep logs while a book is actually in the hands of a patron (think: IP address is assigned by DHCP), and discard the identifying information as soon as the book is returned, or paid for if lost.
Non-identifying information, like "book x circulated twice this year", is retained for planning and statistical purposes.
If one happens to do business in a jurisdiction that has such a requirement, which you can probably discover from the ALA, then you have a perfect right to obey the law and discard old logs once appropriate billing information is obtained from them, or not retain them at all if you do not need them for a legitimate business purpose.
Of course, you will face the same pressures that librarians do in their everyday work (;-))
... then you probably will want to obey U.S. law, especially when it's something a reasonable as the A.D.A.: I wish we had as strong a law up here in Canada.
--dave
It's a public good first, that to, as the Americans say
"promote the Progress of Science and useful Arts", is made into a commodity for a limited time.
To suggest it is a commodity or private good by nature is to fall
into the same trap as to say a Crown or Constitutional grant of privilege is "intellectual property".
--dave
i)ave wrote:Particularly in Germany, there is a sense that the Government is there to protect people from invasion of people's rights by the Corporations
Interesting: in the years leading up to the 2nd World War,
both German and Italy were experimenting with giving the
corporations more say in government, with representatives from,
for example, the oil and gas industry, elected by their companies to committees officially advising on the creation of legislation. This was
formally called "corporatism". In Italy, corporatism was conventionally called "fascism", the name of the main corporatist party.
These same corporation were later castigated for their gleeful support of going to war, in both the first and second world wars,
and for the governments with which they associated.
You might imagine the Italians and especially the Germans would distrust anything and anyone who contributed to their hell-ride into WWII.
Back when new root domains were first proposed, one of
the best proposals to deal with the trade-name-clash
problem was to use the business type, taken from a common
list maintained by the WTO.
That would have make then utube.manuf, which would tend
to reduce the likelyhood of error.
Earthquakes produce slow pressing, shifting and twisting forces, while explosions produce very very rapid pressure rises. The latter are quite a bit better at breaking glass than the former.
Slashdot Mirror
The product, when introduced, was indeed clunky, but has
quietly and steadily improved over the years. I'm a
happy customer, having bought it twice for different
customers.
--dave
The British are trying this as we speak.
--dave
Amusingly, one of the best suggestions for a new boss of some unpopular sex/colour/whatever is to hunt down the departmental "wally" and fire them.
Causes folks to sit up and pay attention when a supposedly "nice" female Canadian fires a lying bastard of a salesman (;-))
--dave
Because MS have to provide backwards compatability for their older releases, the Samba team have the usual 10 years to learn the protocol varient.
This is the same thing that kept IBM from defeating the plug-compatable vendors.
--dave
--dave
You control the app, so you can easily log the response time of the database, and measure it with a small but credible test load.
If you send the database a hundred or so requests at a rate that's deliberately slow, you will get an average response time that's a good estimate of the actual internal response time of the database. Let's say it's 1/10 second. With that number (alone) you can predict and plan for the performance you need.
On a uniprocessor, you will get a maximum of 10 transactions per second before a queue starts to build and you start seeing delays due to queuing, so for 1 through 10 TPS, the response time will be 1/10 second.
In fact, as you get closer and closer to 10 TPS, there is a higher and higher probability that two transactions will come in at the same time, and the queue will start to build before you hit 10...
After the queue builds up, every additional request that comes in will need to wait before it get processed. T queue length is calculated using Little's law, Q=XR, where X is offered load and R is response time. A load of 50 requests would yield a queue length of 50* 1/10 = 5, and the average response time at that load would be (40*0.1 + 0.1) = 4.1 seconds
Voila! For a load of 50 TPS and a target time of 1/10 second, you need a five-processor system, easily achieved with 3 dual-core AMDs.
Feel free to send me mail: I have a copy of Neil Gunther's "pdq" queue solver and can easily compute what-ifs based on your measurements and needs. See "Analyzing Computer System Performance with Perl::PDQ", at http://www.perfdynamics.com/iBook/ppa_new.html
--dave
Normal computer users aren't being served well by Windows, either. Dumbing down an interface won't help my Grandma (actually a friend's grandma) at all: she's a retired MD, businesswoman and druggist. Which is to say she's a normal person, not a moron.
What she needs is a way to change something without writing a program, something she doesn't know how to do. Therefor she needs a graphical representation of the subject matter, such as her patient records, and a way to manipulate and edit it symbolically.
As it happens, the company that did the research in that area is Xerox, and the one which popularized is is Apple. Windows is a non-starter, so don't make the mistake of copying it...
--dave
No, they just wanted something that could implement "time machine", their backup-done-right proposal.
I expect Solaris to be switchd to GPL some time in the GPLv3 era, at which point there won't be a problem porting ZFS to Linux. Not that it was technically difficult to port it to Apple/BSD (;-))
--dave
Simple, true, but hard (;-))
As it happens, there are an amazing number of bad jobs out there, and a moderate number of good ones. One often has to tell your boss that you're conflicted but won't lave hime in the lurch.
Then you can start a job-hunt for a good job, and on finding one, negotiate a reasonable notice period and hand-over.
-dave
Sun's HPC contribution is in optical chip interconnects, described (somewaht fluffily) at http://research.sun.com/spotlight/2006/2006-04-07_ Sun_on_HPCS.html
My Smarter Colleagues noticed that from the same data structure we used for the lotus menus we could build PF-key menus, modern cascading drop-down menus and right-mouse-button pop-up menus.
Which means that for any menu sequence of head->middle->middle*->tail, you can change the visual appearance of the menu without changing the application-level calls used to create it. And that in turn means you can make "ribbon menus" a user-specifiable "skin".
--dave
My parents were advised to stop their children from reading and get them out into the fresh air to play. Reading is unhealthy, you understand (;-))
Mind you, they had to read that advice in a newspaper, requiring them to be readers.
Eventually they heard the same thing about television, both from the newspapers and while sitting on the couch watching the nightly news.
And this week it's games... I wonder if they'll hear about them by email?
--dave
I used to work in a large international firm based in Germany, developing one of their large-scale business offerings.
Some portions used open source software, notably for interoperability with PCs, and we had to say so in the customer documentation.
Customers (and therefor our managers (;-)) wanted three things:
If they had these three, the level of concern about the kind of software it was fell very rapidly. If we only offered them two of the three, they tended to to be very dubious about the whole deal. You can imagine how that motivated us (;-))
As to support, the customers were very open to our company having a relationship with the software developers, with us reporting bugs and proposing patches. This in turn made our managers see the GPL as part of a quite normal business relationship with the developers, and in some cases had us offering the support for the open source components as part of our own service contract.
--dave
The Multicians had a set of rules that allowed them to do "continuous maintenance" on production systems without forcing everything to change at once.
The compatibility didn't last forever, but if you were an application developer, you really only need to do a QA run once a quarter to see if anything was in the process of changing, and schedule a fix for the next quarter. Your customers had a longer guarantee, probably a year or so.
The idea was a lot like relational database theory: you can always "add a column to the table" and you could use NULLs to mark an old column as not containing anything any more, although the actual technology was major- and minor-version-numbers, and was derived from some hardware-versioning research at MIT.
They still work: I used the same techniques in a Unix project, and never had to have a flag day, although Edsel and I were changeling the interface in question with wild abandon (;-))
--dave
This seems to have changed fairly recently: I suspect the license-challenged are researching what part is pure BSD by now (;-))
--dave
Actually, quite a number of folks who are my consulting customers use Darwin (really BSD) sources as the "reference copies" of programs they're adapting for their own use.
This is in part because of the good quality of the code, and the company which stands behind it. In part it is because of the larger BSD community who stands semi-invisibly behind Apple... some customers really understand the strength of community. And finally, for the license-paranoid, in part this is because of the use of the very old and weak BSD license.... some customers really don't understand the community (;-))
Coming back to the main point of the discussion, adoption of the GPL by well-known fortune-500 companies is a step away from the world of Microsoft, SCO and FUD.
Definitely a change, and definitely for the better.
--dave
In a previous life, I had a smart-card for a badge, which I shoved in a sunray x-terminal or a laptop as the"thing I had", and typed a password as the "thing I knew", after which I got my current session back.
If I needed to so somewhere else, I unplugged the card and my session was saved. When I got there, I plugged back in again, typed my password to the screen-saver and picked up exactly where I left off.
I was very pleased with this scheme: it saved me hours of frustration with AD kludgery and the string of crypto-keyfobs I now have to cart around.
--dave
And it's GPL, same as Open Office (from many moons ago).
I suspect you'll see more of this: Sun uses specialized licenses only when there is some legal reason to do so, where "legal reason" can mean "Microsoft fork", "patent trap" or something less contentious such as "we don't own that code, so we can't sublicense it".
--dave
... in enough english-speaking jurisdictions in North America that library software companies arrange for their programs to only keep logs while a book is actually in the hands of a patron (think: IP address is assigned by DHCP), and discard the identifying information as soon as the book is returned, or paid for if lost.
Non-identifying information, like "book x circulated twice this year", is retained for planning and statistical purposes.
If one happens to do business in a jurisdiction that has such a requirement, which you can probably discover from the ALA, then you have a perfect right to obey the law and discard old logs once appropriate billing information is obtained from them, or not retain them at all if you do not need them for a legitimate business purpose.
Of course, you will face the same pressures that librarians do in their everyday work (;-))
--dave
... then you probably will want to obey U.S. law, especially when it's something a reasonable as the A.D.A.: I wish we had as strong a law up here in Canada.
--dave
It's a public good first, that to, as the Americans say "promote the Progress of Science and useful Arts", is made into a commodity for a limited time.
To suggest it is a commodity or private good by nature is to fall into the same trap as to say a Crown or Constitutional grant of privilege is "intellectual property".
--dave
i)ave wrote:Particularly in Germany, there is a sense that the Government is there to protect people from invasion of people's rights by the Corporations
Interesting: in the years leading up to the 2nd World War, both German and Italy were experimenting with giving the corporations more say in government, with representatives from, for example, the oil and gas industry, elected by their companies to committees officially advising on the creation of legislation. This was formally called "corporatism". In Italy, corporatism was conventionally called "fascism", the name of the main corporatist party.
These same corporation were later castigated for their gleeful support of going to war, in both the first and second world wars, and for the governments with which they associated.
You might imagine the Italians and especially the Germans would distrust anything and anyone who contributed to their hell-ride into WWII.
--dave
--dave
Back when new root domains were first proposed, one of the best proposals to deal with the trade-name-clash problem was to use the business type, taken from a common list maintained by the WTO.
That would have make then utube.manuf, which would tend to reduce the likelyhood of error.
--dave
Earthquakes produce slow pressing, shifting and twisting forces, while explosions produce very very rapid pressure rises. The latter are quite a bit better at breaking glass than the former.