.NET/C# versus PHP

[Jerry Stuckle]

Even if they will only make 90% or 95% it is still good.

Prove they are even near 90 or 95%. If the data are accurate, it is
provable.

No.

They claim to measure sites and they measure sites.

That is not being off.

But they can't prove their claims.

Not at all.

Netcraft show what they claim to show.

Arne

Yes, they show what they "claim" to show. That does not make it accurate.

I can interview 100 million people in the United States. That's a huge
number, so it must be accurate! And 95% of them claim they are male.

So my conclusion is that 95% of the people in the United States are male.

This, according to you, is a valid conclusion.

Sorry - you need to study up on statistics, sampling and even just how
many sites are out there.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

Never mind. I did a quick search on PHP and Python mailing lists, and
realized that my post managed to push all your buttons at once - it
looks like there's at least one "PHP vs ..." thread you participate in
regularly, wherever it appears, and the arguments from the old threads
are all oddly familiar as well. I guess this one is best left to rest.
Well, except for this one bit, which should be corrected for factual
reasons (so that no poor soul stumbles onto this thread and gets a
wrong idea):


No, they do not. When you run VBScript using its interpreter (e.g.
Windows Scripting Host), the .vbs file does not have any <%
characters, and, indeed, were they to be there, it would be a
syntactic error. Nor are they present in the code when you use
VBScript inside HTML <script> tag. They are ASP (framework /
templating engine) artifacts, not language ones - if you use e.g.
PerlScript from ActivePerl with ASP, then you'll get them in Perl,
too. But of course they aren't part of the language.


They aren't part of Java _language_ either. They are definitely part
of JSP, which is also a framework / templating engine. They are parsed
by JSP parser, and <% %> symbols aren't part of the Java scriptlet
itself, they merely indicate its boundaries.

JSP is not a framework/templating engine. You obviously are not at all
familiar with what the term means.

The difference between all those examples and PHP is that for PHP, the
<?php ?> syntax is the core part of the language itself, and _every_
PHP source file is treated as text with code islands (with code
islands being ones that have to be explicitly marked), whether it
makes sense or not (which it usually doesn't).

Yea, when all you have is a hammer, every problem is a nail.

Fortunately, PHP gives you lots of tools to work with. You aren't
restricted to the hammer.

But so long. I've learned long ago not to try to have an intelligent
discussion with folks like you. It's impossible.


--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Michael Austin]

I thought WebObjects were Java ??


SUN's Java implementation is GPL for tools and JVM but GPL
with classpath exceptions for the libraries.


PHP is very much bazaar not cathedral.

Arne

Yes there is JAVA - but, getting knowledgeable JAVA developers and
support staff that understands the beast and can tune it PROPERLY, can
be a daunting task in and of itself! (And that AIN'T ORACLE SUPPORT!!!)
We had a Oracle BPEL process that was killing us - turned out it was
taking over 60 seconds to do GC every 5 minutes or so - granted the
thing was 25GB - and Oracle was recommending increasing it to 35GB (box
had 192GB so size was not an issue). By the time we got finished with
it (and using an internal resource that was very good at Java and GC
tuning, got it down to ~10GB and throughput increased by 4-500%.

 

[Arne Vajhøj]

Yes there is JAVA - but, getting knowledgeable JAVA developers and
support staff that understands the beast and can tune it PROPERLY, can
be a daunting task in and of itself! (And that AIN'T ORACLE SUPPORT!!!)
We had a Oracle BPEL process that was killing us - turned out it was
taking over 60 seconds to do GC every 5 minutes or so - granted the
thing was 25GB - and Oracle was recommending increasing it to 35GB (box
had 192GB so size was not an issue). By the time we got finished with
it (and using an internal resource that was very good at Java and GC
tuning, got it down to ~10GB and throughput increased by 4-500%.

If you want to get a serious server running optimal then hiring
someone with knowledge about the server is generally a good
thing.

Large Oracle database, big Java EE application server or something
else does not matter.

Arne

 

[Michael B. Trausch]

I didn't say whether the sample was biased or not. I said it was
scientifically invalid - which it is. Therefore, the results are
invalid.

Whether there is a "more accurate" sampling or not is also
immaterial. Lack of something "more accurate" does not make this
accurate.

No, but unless you know of a better source of data, this is the best
there is. The value is not zero, unless you are a moron.

--- Mike

 

[Michael B. Trausch]

Sorry, Arne, the data are 100% meaningless. It is not a scientific
sample. Therefore, results are garbage.

I don't understand what's so hard for you to understand, Jerry. You're
essentially claiming that the data's net worth is nil, nothing. This
is _hardly_ the case. Even incomplete data is better than no data at
all, and any one who practices competently in any field of science will
tell you that.

The data are perhaps meaningless to _you_, but there are really no
valid reasons to reject the data: yes, it is incomplete. But the data
_over time_ is the interesting part of the Netcraft surveys. If you
fail to see that, well, that's a personal problem. Any statements
regarding your subjective view are just that, anyway: subjective. So,
you don't care about the data. Fine. But that does not necessitate
any need to trash-talk the data.

Given that the samples are incomplete, and _everyone_ knows this, that
has to be considered when doing anything (if you're doing anything)
based on that data. All that the data tells you is that there are _at
least_ X number of public domains using Y number of Web servers, and it
provides a loose order of ranking for the known Web servers out there.
The data's quality constantly improves with the addition of new hosts,
and therefore necessarily becomes more accurate over time. And yet,
trends are fairly consistent. This speaks volumes, whether or not you
want to state that its "garbage" or not.

What the data won't (and _can't_) tell you is how popular a given
language environment for Web programming is. I run ASP.NET
applications (and PHP applications, and Python applications) on my Linux
server (on Apache) at home, for example. Various clients of mine use a
combination of those languages, as well as others. The type of Web
server, and even the type of operating system, is quite irrelevant when
talking about the environment that it provides. What _is_ interesting
is the relative popularity of certain choices of software. If you've
no use for that data, then fine. But that in itself doesn't make it
"meaningless," nor does it make it "garbage".

--- Mike

 

[Michael B. Trausch]

Also, a survey does not need to be complete to be accurate. But it
does need a representative sample - which this doesn't have.

Perhaps you have a method for generating a representative sample that
is statistically accurate on the Internet that you can employ and get
better quality data?

If so, do it. You'll make money.

--- Mike

 

[Michael B. Trausch]

Yes, they show what they "claim" to show. That does not make it
accurate.

I can interview 100 million people in the United States. That's a
huge number, so it must be accurate! And 95% of them claim they are
male.

So my conclusion is that 95% of the people in the United States are
male.

This, according to you, is a valid conclusion.

If that's what you assume you can do with _any_ statistic, you're more
of an idiot than I'd previously thought. Even well done, controlled
experiments that require a "representative sample" aren't accurate.
They _can't_ be accurate, by definition. "Representative sample" is
only meaningful in a very small way, because any such sample is going
to be what one person or group _thinks_ is representative.

After all, you can't _accurately_ determine what a real representative
subset of anything is unless you first actually have data on the whole
set that the subset claims to represent. If you can do that, you are
likely psychic. Or, more likely, just an ego-inflated idiot.

The fact that NetCraft says that X% of servers they have polled are
running Apache does _not_ mean that X% of servers on the _INTERNET_ are
running Apache. It means exactly what it says and not a damn thing
more.

--- Mike

 

[Michael B. Trausch]

Which platforms run Mono but not Java?

Mono runs on every platform that is compliant with POSIX. It doesn't
have a JIT on every such platform, but it can be run with an
interpreter on every such platform. Mono's only dependency outside of
POSIX is GLib.

Java, OTOH, is available in whatever binaries they give you on the Java
Web site for the "real deal". There are other implementations of the
JVM, but none of them that I have actually encountered implement enough
of Java to be worthwhile using. Many of those projects have died over
the years, as well, and no longer build on current systems (and in some
cases, were left in such a deficient state that they don't work on the
ones they were intended for, either). There are alternative JVMs out
there that will do 60% or maybe 70% of what you need, but it doesn't
take long to find a Java program or applet that won't run correctly, if
at all, within them.

OTOH, Mono 2.4 runs literally everything I have thrown at it, and can
be built for a great number of systems and function identically on
every one of them. It helps that Mono is completely free software.
Java isn't (yet, though it should be Real Soon Now(tm), perhaps then
it will also gain in compatible portability).

Mono also, AFAICT, implements most if not all of the standard
specification, which is something that alternative JVMs cannot do,
since there is no standard other than the de facto one.

--- Mike

 

[Jerry Stuckle]

No, but unless you know of a better source of data, this is the best
there is. The value is not zero, unless you are a moron.

--- Mike

I never said a value is zero or ANYTHING else. I only said the data are
invalid.

And bad data is WORSE than no data - you don't come to incorrect
conclusions from no data.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

I don't understand what's so hard for you to understand, Jerry. You're
essentially claiming that the data's net worth is nil, nothing. This
is _hardly_ the case. Even incomplete data is better than no data at
all, and any one who practices competently in any field of science will
tell you that.

Anyone competent in any field will tell you. Invalid data is worse than
no data at all. That is why polls state a margin of error. They can do
that because their sample is selected using scientific principals to
ensure they get a valid cross-section. This uses no such selection
criteria; the possible margin of error is so huge the results are
meaningless - as anyone competent in any field will tell you.

The data are perhaps meaningless to _you_, but there are really no
valid reasons to reject the data: yes, it is incomplete. But the data
_over time_ is the interesting part of the Netcraft surveys. If you
fail to see that, well, that's a personal problem. Any statements
regarding your subjective view are just that, anyway: subjective. So,
you don't care about the data. Fine. But that does not necessitate
any need to trash-talk the data.

There is no valid reason to accept the data. It is not that the data
are incomplete; it's the selection process to get the data.

I could generate data which shows 99% of servers are Apache and 1% are
IIS. I could generate data which shows just the opposite. Either set
of data would be just as valid as this data. What makes that so hard
for you to understand?

Given that the samples are incomplete, and _everyone_ knows this, that
has to be considered when doing anything (if you're doing anything)
based on that data. All that the data tells you is that there are _at
least_ X number of public domains using Y number of Web servers, and it
provides a loose order of ranking for the known Web servers out there.
The data's quality constantly improves with the addition of new hosts,
and therefore necessarily becomes more accurate over time. And yet,
trends are fairly consistent. This speaks volumes, whether or not you
want to state that its "garbage" or not.

It doesn't even show you that. All it shows you is that over whatever
period of time the sample was taken, these servers were detected. How
many of these sites are still active, and how many are still using the
same web server, for instance?

As to the loose order of ranking - not even that. Let's say I take an
poll of people in America. It just so happens 95% of my interviewees
are men. So according to you, I should come to the conclusion that
about 95% of the people in the United States are men, and that men by
far outnumber women - neither of which is true. However, according to
your arguments, these statistics are valid.

What the data won't (and _can't_) tell you is how popular a given
language environment for Web programming is. I run ASP.NET
applications (and PHP applications, and Python applications) on my Linux
server (on Apache) at home, for example. Various clients of mine use a
combination of those languages, as well as others. The type of Web
server, and even the type of operating system, is quite irrelevant when
talking about the environment that it provides. What _is_ interesting
is the relative popularity of certain choices of software. If you've
no use for that data, then fine. But that in itself doesn't make it
"meaningless," nor does it make it "garbage".

--- Mike

What you can't tell is ANYTHING valid.

I suggest you study statistics and how to get accurate ones. It's been
35 years since I had college classes in them, but I'm sure things
haven't changed that much.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

Perhaps you have a method for generating a representative sample that
is statistically accurate on the Internet that you can employ and get
better quality data?

If so, do it. You'll make money.

--- Mike

That's a whole study in itself, Mike. But it can be done - companies do
it all the time. But it can't be done by taking a bunch of random sites
like netcraft has done.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

If that's what you assume you can do with _any_ statistic, you're more
of an idiot than I'd previously thought. Even well done, controlled
experiments that require a "representative sample" aren't accurate.
They _can't_ be accurate, by definition. "Representative sample" is
only meaningful in a very small way, because any such sample is going
to be what one person or group _thinks_ is representative.

That is true. But that representative sample is selected is selected
via proven criteria, and is generally accepted as valid by others.

After all, you can't _accurately_ determine what a real representative
subset of anything is unless you first actually have data on the whole
set that the subset claims to represent. If you can do that, you are
likely psychic. Or, more likely, just an ego-inflated idiot.

Incorrect.

The fact that NetCraft says that X% of servers they have polled are
running Apache does _not_ mean that X% of servers on the _INTERNET_ are
running Apache. It means exactly what it says and not a damn thing
more.

--- Mike

Their selection criteria is flawed, and so is their data. It doesn't
mean anything that you or others here have claimed.

You really need to study statistics. There are ways to come up with a
scientifically valid representative sample. It is done every day. And
it is accepted by experts in every field where it is done.

But netcraft's way is not the way to do it.

And the fact you are now resorting to the ad hominem attacks proves you
have no valid argument. Such is the case with trolls.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Michael B. Trausch]

I never said a value is zero or ANYTHING else. I only said the data
are invalid.

"Invalid" must mean something very different to you than what I
understand it to mean.

And bad data is WORSE than no data - you don't come to incorrect
conclusions from no data.

But that's just it, the data _isn't_ bad. It's _partial_.

--- Mike

 

[Jerry Stuckle]

"Invalid" must mean something very different to you than what I
understand it to mean.


But that's just it, the data _isn't_ bad. It's _partial_.

--- Mike

Partial does not mean good. Data not scientifically sampled is garbage.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

I am not arguing with you Jerry, just seeking information, as I am not
familiar with Netcraft:

What, specifically, in your opinion do you find fault with in the manner
in which Netcraft selected their sample and collected their data?

They have followed none of the basic principles in selecting and
validating their data. They cannot show a probability of error because
their data was not properly acquired.

Sample selection for statistical analysis is a science all into itself,
and one which cannot be performed accurately by randomly picking web
sties (or other samples).

- Is it that the sample population is biased, and hence not
representative of the general population?
- Is it with the sampling techniques, once the sample was chosen?
- Is the the type data that were collected once their sample was
determined and the method of collecting data was determined?

It can be all of the above. Their failure to follow any basic sampling
principles invalidates their data.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Jerry Stuckle]

Yes, I know. I taught it, in fact.


You make a rule. Say, select every 10th to come up in a google search
based upon a random selection from a list of 1000 words. That is a
rule, and the selection is random. The results will then be as good as
the 1000 words samples the population space.

So, are you using "random" here in the colloquial or statistical sense?

But to get valid data on all web sites, you can't just come up with any
old word and pick. That's especially true when you're dealing with
sites written in hundreds of languages around the world.

As a trivial example - pick "microsoft" - the results are most likely
going to be heavily weighted in one obvious direction. Pick "Linux" and
just the opposite will be true. Pick "three" and you're going to leave
out a lot of non-English sites. The list goes on and on.

And I'm using "random" in the statistical sense. Just picking a word
does not give a statistically random sampling.

It depends upon their sample state vs the population space and the rules
they choose for sampling.

That's part of it, yes. But even with 200M web sites, they are no where
near any significant percentage of the number of websites in the world.


--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Michael Austin]

If you want to get a serious server running optimal then hiring
someone with knowledge about the server is generally a good
thing.

Large Oracle database, big Java EE application server or something
else does not matter.

Arne

I concur, except when they do really stupid things like "lay off" those
really smart resources "to get their cost in line with the competition".

Typical conversation seems to be...
"Hmmm we have 10 people supporting this at tier one - off-shore. We
have 1 on-shore that is really good at this tuning - hey I know - let's
get rid of the one on-shore guy and hire 3 more offshore to take his
place". (meanwhile performance is in the toilet and they still can't
figure out why...)

 

[Jerry Stuckle]

You are right, and that is why I added "The results will then be as good
as the 1000 words samples the population space.". The example I gave
was just that, an example, and was off the top of my head. As you
illustrate, it is not a good method. The real question I was asking was
what sampling criterion did they use and how well does it mirror the
population space. Based upon your previous replies you seem to be
saying that they didn't have a rule and just tried to grab whatever they
could.

That's how they do it.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[Erwin Moller]

Pavel Minaev schreef:
Hi Pavel,

First of all, I would like to remind you that the world isn't limited
by PHP and ASP.NET.

No need to remind me of that. ;-)

There's also Java, there's Perl, Python, and Ruby,

and there are many more exotic solutions (such as Apple's WebObjects).

Indeed. I tried them all (except Ruby) to different extends.

Did it actually ever help you (i.e. did you, in the course of web
application development, had to patch PHP source code)?

Then, of course, Java is GPL'd, and Python and Ruby are under BSDL-
like licenses as well.

Yes, but the OP was comparing .NET/C# with PHP.

This is very subjective. I can accept that PHP is easier to learn
than .NET or Java, but I seriously doubt that it is easier to learn
than Python.

agree.
But again: The OP was mainly interested in .NET and PHP.

Again, this omits Python and Ruby. The latter is particularly
extremely expressive, the former is more verbose, but many find it to
be a golden standard of readability.

But, of course, this whole point is largely subjective.

With respect to "learning enormous amounts of classes" - this is
equally true about PHP, but there it's "learning enormous amounts of
functions". Which is made far worse by the fact that those functions
are rather badly organized (thanks to the lack of namespaces, among
other things), and very inconsistent - you get stuff like, in the same
family of functions, the order of arguments is occasionally reversed
for no obvious reason.

Well, I did both languages a lot, and don't agree here.
But I DO agree with you this is all very subjective.

Yes, of course one must learn some basic functions in PHP. But they are
easily found and reveived, and (most of the time) logically named.
In Java one must learn a lot more before getting on speed.
Allthough in JavaFX things are made simplier I understand, but I haven't
tried that seriously yet.

All of these points equally well apply to Perl, Python, or Ruby,
except that all three are much better at "coding structured" than PHP.

Yes, we agree on that.
It is my impression the OP was mainly interested in PHP versus .NET.

This is the only one that PHP can actually claim for itself
exclusively

Regards,
Erwin Moller

--
"There are two ways of constructing a software design: One way is to
make it so simple that there are obviously no deficiencies, and the
other way is to make it so complicated that there are no obvious
deficiencies. The first method is far more difficult."
-- C.A.R. Hoare