Re: Why people not using mod_perl

On Tue, 30 Nov 2004 22:38:11 +0000
modperl[at]att.net wrote:

> 3) capacity/scalable
>
> mod_perl is very scalable --- I mean, one can properly
> config a single server to handle dynamic content for
> 200K daily unique IPs. PHP may end up with just 100K
> and servlet ends up at around 50K.
>

I'm just curious, is this performance data still true in today?
We have a new project building a website for a goverment which should
handle lots of transportation data, servlet and modperl are two
choices. So I googled and found this old message.

http://www.gossamer-threads.com/lists/modperl/advocacy/75311

Re: Why people not using mod_perl

On Wed, Sep 16, 2009 at 4:52 AM, Jeff Pang wrote:
> I'm just curious, is this performance data still true in today?
> We have a new project building a website for a goverment which should handle
> lots of transportation data, servlet and modperl are two choices.

I don't know what the source of that data was. However, mod_perl is
basically just Perl, and Perl is very fast. In most of the language
benchmarks I've seen, Perl comes out a little ahead of PHP and
somewhat behind Java. In real-world websites though, Perl often ends
up being faster than Java because of slow Java web frameworks and the
overly-abstract designs they encourage.

You can certainly succeed at building large websites with either Perl
or Java. I'd suggest you consider who will be doing the work and what
the expenses will be. If you decide to use Java, go with open source.
The commercial frameworks are slow and not worth the price.

- Perrin

Re: Why people not using mod_perl

--0015175cac9c7b99250473b89cfd
Content-Type: text/plain; charset=ISO-8859-1

My site algebra.com is about 80,000 lines of mod_perl code.

I wrote a relatively large framework, with many homegrown perl modules,
about five years ago.
It uses a database, image generation modules, a big mathematical engine that
I wrote (that "shows work", unlike popular third party packages), etc.

All pages of my site are dynamic and it is very image heavy due to math
formulae.

I can say two things:

1) It is relatively fast, serving pages in 0.1 seconds or so

2) Despite the quantity of code, and its age, it is still very maintainable
and understandable (to me).

If I was to make a choice again, I would go with mod_perl again. With Perl,
I can "stand on the shoulders of giants" like Lincoln etc, and use the
brilliant stuff they provided to serve my users.

i

--0015175cac9c7b99250473b89cfd
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

My site is about 80,000 line=
s of mod_perl code.

I wrote a relatively large framework, with many =
homegrown perl modules, about five years ago.
It uses a database, image=
generation modules, a big mathematical engine that I wrote (that "sho=
ws work", unlike popular third party packages), etc.


All pages of my site are dynamic and it is very image heavy due to math=
formulae.

I can say two things:

1) It is relatively fast, =
serving pages in 0.1 seconds or so

2) Despite the quantity of code, =
and its age, it is still very maintainable and understandable (to me).


If I was to make a choice again, I would go with mod_perl again. With P=
erl, I can "stand on the shoulders of giants" like Lincoln etc, a=
nd use the brilliant stuff they provided to serve my users.

i


--0015175cac9c7b99250473b89cfd--

Re: Why people not using mod_perl

On Wed, Sep 16, 2009 at 2:30 PM, Igor Chudov wrote:
> My site algebra.com is about 80,000 lines of mod_perl code.
>
> I wrote a relatively large framework, with many homegrown perl modules,
> about five years ago.
> It uses a database, image generation modules, a big mathematical engine that
> I wrote (that "shows work", unlike popular third party packages), etc.
>
> All pages of my site are dynamic and it is very image heavy due to math
> formulae.
>
> I can say two things:
>
> 1) It is relatively fast, serving pages in 0.1 seconds or so

It is fast, just visited it. However, I think you could get a faster
initial page load by compressing the html returned to the client using
mod_deflate. The main page request was 137 milliseconds and you could
probably drop that by 30-50% with gzipping the output.

Nice work!

> 2) Despite the quantity of code, and its age, it is still very maintainable
> and understandable (to me).
>
> If I was to make a choice again, I would go with mod_perl again. With Perl,
> I can "stand on the shoulders of giants" like Lincoln etc, and use the
> brilliant stuff they provided to serve my users.

Yes, that's why I use mod_perl. The core developers on the Apache
Httpd, and mod_perl core are world class.

Re: Why people not using mod_perl

--- On Thu, 9/17/09, Igor Chudov wrote:
> My site algebra.com is about 80,000
> lines of mod_perl code.
>=20
> I wrote a relatively large framework, with many homegrown
> perl modules, about five years ago.=20
> It uses a database, image generation modules, a big
> mathematical engine that I wrote (that "shows
> work", unlike popular third party packages), etc.=20
>=20
>=20
> All pages of my site are dynamic and it is very image heavy
> due to math formulae.=20
>=20
> I can say two things:=20
>=20
> 1) It is relatively fast, serving pages in 0.1 seconds or
> so
>=20
> 2) Despite the quantity of code, and its age, it is still
> very maintainable and understandable (to me).=20

In that case, would you like to fix its mangled output?

e.g. http://www.algebra.com/algebra/homework/divisibility/Prime_f actorizati=
on_algorithm.wikipedia

  (Redirected from Prime factorization algorithm)

faster than O((1+ε)b) for all positive ε

an integer M with 1 â‰Â=A4 M â‰Â=A4 N

Pollard's p âˆâ€=99 1 algorithm

Section 4.5.4: Factoring into Primes, pp. 379–417.

Chapter 5: Exponential Factoring Algorithms, pp. 191â€â=80=
=9C226. Chapter 6: Subexponential Factoring Algorithms, pp. 227ââ‚=
¬â€œ284. Section 7.4: Elliptic curve method, pp. 301ââ‚=
â€=9C313.

Eric W. Weisstein, â€Å=93RSA-640 Factoredâ€Â=


v â€Â=A2 d â€Â=A2 e

AKS · APR · Ballie–PSW Â=
· ECPP · Fermat · Lucas · Lucasâ=
€“Lehmer · Lucas–Lehmer=
–Riesel · Proth's theorem · P=
épin's · Solovay–Strassen =C3=
‚Â=B7 Miller–Rabin · Trial division

Sieve of Atkin · Sieve of Eratosthenes · Sieve of Sun=
daram · Wheel factorization

CFRAC · Dixon's · ECM · Euler's Â=C2=
=B7 Pollard's rho · P âˆâ€=99 1 · P + 1=
· QS · GNFS · SNFS · rational =
sieve · Fermat's · Shanks' square forms · =
Trial division · Shor's

Ancient Egyptian multiplication · Aryabhata · Binary =
GCD · Chakravala · Euclidean · Extended Eu=
clidean · integer relation algorithm · integer square=
root · Modular exponentiation · Schoof's Â=C2=
=B7 Shanks-Tonelli



Looks like you've got utf8 and iso8859-1 messed up.

Phil

=0A

Re: Why people not using mod_perl

It gets worse:
"""
Factor any number

Find all factors of [305550321722429173]

Solution by Factor any number
305550321722429173 is NOT a prime number: 305550321722429173 = 2 * 2 * 2 * 2 * 2 * 2 * 2 * 2 * 3 * 3 * 3 * 3 * 199 * 293 * 252718517
Work Shown

305550321722429173 is divisible by 2: 305550321722429173 = 1.52775160861215e+17 * 2.
1.52775160861215e+17 is divisible by 2: 1.52775160861215e+17 = 7.63875804306073e+16 * 2.
7.63875804306073e+16 is divisible by 2: 7.63875804306073e+16 = 3.81937902153036e+16 * 2.
3.81937902153036e+16 is divisible by 2: 3.81937902153036e+16 = 1.90968951076518e+16 * 2.
1.90968951076518e+16 is divisible by 2: 1.90968951076518e+16 = 9.54844755382591e+15 * 2.
9.54844755382591e+15 is divisible by 2: 9.54844755382591e+15 = 4.77422377691296e+15 * 2.
4.77422377691296e+15 is divisible by 2: 4.77422377691296e+15 = 2.38711188845648e+15 * 2.
2.38711188845648e+15 is divisible by 2: 2.38711188845648e+15 = 1.19355594422824e+15 * 2.
1.19355594422824e+15 is divisible by 3: 1.19355594422824e+15 = 397851981409413 * 3.
397851981409413 is divisible by 3: 397851981409413 = 132617327136471 * 3.
132617327136471 is divisible by 3: 132617327136471 = 44205775712157 * 3.
44205775712157 is divisible by 3: 44205775712157 = 14735258570719 * 3.
14735258570719 is divisible by 199: 14735258570719 = 74046525481 * 199.
74046525481 is divisible by 293: 74046525481 = 252718517 * 293.
252718517 is not divisible by anything.
"""

Re: Why people not using mod_perl

This is a multi-part message in MIME format.
--------------080104040908000800010602
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit



This is a mod_perl list, so I would expect to see Perl championed pretty
heavily, but Java, .net and there ilk are undoubtedly *the* choice for
large web applications. I'd like to get into some discussion as to why
almost all *large* sites choose these languages.

I don't have any experience developing a large application in Java,
although I do have a lot of experience working on the operations side of
a large web application that is Java based.

The reasons I generally hear for choosing Java over mod_perl are:

1) Speed - I don't buy this at all
2) Maintainability - I think this makes sense. Perl can be pretty easy
to maintain if you stick a good framework around it, but you have to
seek out that framework and YOU are responsible for adhereing to it.
All of that is inherent in Java. It also helps that Java has OO built in.
3) Easier to package and build/move code - In my experience this is true.
4) Advantages to be gained from running on an actually application
server - Also valid
5) Compatible enterprise class middleware - Also true, Java plugs into
more truly enterprise level suff than Perl does. (security frameworks,
etc... )
6) Support

A lot of the industry seems look at Perl as obsolete technology that has
been replaced by *insert hot new technology of the week here* which is
a total shame. I've worked with a lot of technologies and I think Perl
is a great choice for small/medium websites and webapps, which is
probably what most of us work on. But I'm very interested to know at
what point (if any) a site/app grows too large or too complex for
mod_perl and what defines that turning point. Could Amazon run on
mod_perl for example?
>
>
>
>
> Phil Carmody wrote:
>> --- On Thu, 9/17/09, Igor Chudov wrote:
>>
>>> My site algebra.com is about 80,000
>>> lines of mod_perl code.
>>>
>>> I wrote a relatively large framework, with many homegrown
>>> perl modules, about five years ago.
>>> It uses a database, image generation modules, a big
>>> mathematical engine that I wrote (that "shows
>>> work", unlike popular third party packages), etc.
>>>
>>>
>>> All pages of my site are dynamic and it is very image heavy
>>> due to math formulae.
>>>
>>> I can say two things:
>>>
>>> 1) It is relatively fast, serving pages in 0.1 seconds or
>>> so
>>>
>>> 2) Despite the quantity of code, and its age, it is still
>>> very maintainable and understandable (to me).
>>>
>>
>> In that case, would you like to fix its mangled output?
>>
>> e.g. http://www.algebra.com/algebra/homework/divisibility/Prime_f actorization_algorithm.wikipedia
>>
>> Â Â (Redirected from Prime factorization algorithm)
>>
>> faster than O((1+ε)b) for all positive ε
>>
>> an integer M with 1 ≤ M ≤ N
>>
>> Pollard's p − 1 algorithm
>>
>> Section 4.5.4: Factoring into Primes, pp. 379–417.
>>
>> Chapter 5: Exponential Factoring Algorithms, pp. 191–226. Chapter 6: Subexponential Factoring Algorithms, pp. 227–284. Section 7.4: Elliptic curve method, pp. 301–313.
>>
>> Eric W. Weisstein, “RSA-640 Factored”
>>
>> v • d • e
>>
>> AKS · APR · Ballie–PSW · ECPP · Fermat · Lucas · Lucas–Lehmer · Lucas–Lehmer–Riesel · Proth's theorem · Pépin's · Solovay–Strassen · Miller–Rabin · Trial division
>>
>> Sieve of Atkin · Sieve of Eratosthenes · Sieve of Sundaram · Wheel factorization
>>
>> CFRAC · Dixon's · ECM · Euler's · Pollard's rho · P − 1 · P + 1 · QS · GNFS · SNFS · rational sieve · Fermat's · Shanks' square forms · Trial division · Shor's
>>
>> Ancient Egyptian multiplication · Aryabhata · Binary GCD · Chakravala · Euclidean · Extended Euclidean · integer relation algorithm · integer square root · Modular exponentiation · Schoof's · Shanks-Tonelli
>>
>>
>>
>> Looks like you've got utf8 and iso8859-1 messed up.
>>
>> Phil
>>
>>
>>
>>
>>
>>

--------------080104040908000800010602
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit











This is a mod_perl list, so I would expect to see Perl championed
pretty heavily, but Java, .net and there ilk are undoubtedly *the*
choice for large web applications.  I'd like to get into some
discussion as to why almost all *large* sites choose these languages.



I don't have any experience developing a large application in Java,
although I do have a lot of experience working on the operations side
of a large web application that is Java based.



The reasons I generally hear for choosing Java over mod_perl are:



1) Speed - I don't buy this at all

2) Maintainability - I think this makes sense.  Perl can be pretty easy
to maintain if you stick a good framework around it, but you have to
seek out that framework and YOU are responsible for adhereing to it. 
All of that is inherent in Java.  It also helps that Java has OO built
in. 

3) Easier to package and build/move code - In my experience this is
true.

4) Advantages to be gained from running on an actually application
server - Also valid

5) Compatible enterprise class middleware - Also true, Java plugs into
more truly enterprise level suff than Perl does. (security frameworks,
etc... ) 

6) Support



A lot of the industry seems look at Perl as obsolete technology that
has been replaced by *insert hot new technology of the week here* 
which is a total shame.  I've worked with a lot of technologies and I
think Perl is a great choice for small/medium websites and webapps,
which is probably what most of us work on.  But I'm very interested to
know at what point (if any) a site/app grows too large or too complex
for mod_perl and what defines that turning point.   Could Amazon run on
mod_perl for example?

Phil Carmody wrote:

type="cite">

--- On Thu, 9/17/09, Igor Chudov 
 moz-do-not-send="true" class="moz-txt-link-rfc2396E"

 href="mailto:ichudov@gmail.com"><ichudov@gmail.com> wrote:

  

My site algebra.com is about 80,000

lines of mod_perl code.



I wrote a relatively large framework, with many homegrown

perl modules, about five years ago. 

It uses a database, image generation modules, a big

mathematical engine that I wrote (that "shows

work", unlike popular third party packages), etc. 





All pages of my site are dynamic and it is very image heavy

due to math formulae. 



I can say two things: 



1) It is relatively fast, serving pages in 0.1 seconds or

so



2) Despite the quantity of code, and its age, it is still

very maintainable and understandable (to me). 

    

In that case, would you like to fix its mangled output?



e.g. 
  href="http://www.algebra.com/algebra/homework/divisibility/P rime_factorization_algorithm.wikipedia">http://www.algebra.c om/algebra/homework/divisibility/Prime_factorization_algorit hm.wikipedia



  (Redirected from Prime factorization algorithm)



faster than O((1+ε)b) for all positive ε



an integer M with 1 ≤ M ≤ N



Pollard's p − 1 algorithm



Section 4.5.4: Factoring into Primes, pp. 379–417.



Chapter 5: Exponential Factoring Algorithms, pp. 191–226. Chapter 6: Subexponential Factoring Algorithms, pp. 227–284. Section 7.4: Elliptic curve method, pp. 301–313.



Eric W. Weisstein, “RSA-640 Factored” 



v • d • e



AKS · APR · Ballie–PSW · ECPP · Fermat · Lucas · Lucas–Lehmer · Lucas–Lehmer–Riesel · Proth's theorem · Pépin's · Solovay–Strassen · Miller–Rabin · Trial division



Sieve of Atkin · Sieve of Eratosthenes · Sieve of Sundaram · Wheel factorization



CFRAC · Dixon's · ECM · Euler's · Pollard's rho · P − 1 · P + 1 · QS · GNFS · SNFS · rational sieve · Fermat's · Shanks' square forms · Trial division · Shor's



Ancient Egyptian multiplication · Aryabhata · Binary GCD · Chakravala · Euclidean · Extended Euclidean · integer relation algorithm · integer square root · Modular exponentiation · Schoof's · Shanks-Tonelli







Looks like you've got utf8 and iso8859-1 messed up.



Phil









      

  

Re: Why people not using mod_perl

On Thu, Sep 17, 2009 at 2:32 AM, Perrin Harkins wrote:

>
> I don't know what the source of that data was. =A0However, mod_perl is
> basically just Perl, and Perl is very fast.


I think the more exact statement should be, mod_perl is compiled perl,
mod_perl is very fast.
But perl CGI...I must say it's very slow.

Jenn.

Re: Why people not using mod_perl

On Wed, Sep 16, 2009 at 9:42 PM, Jenn G. wrote:
> I think the more exact statement should be, mod_perl is compiled perl,
> mod_perl is very fast.
> But perl CGI...I must say it's very slow.

Well, you can say CGI is slow, but Perl CGI is very fast compared to
the alternatives. Have you ever tried Java CGI? Or PHP CGI? They're
not fast.

Also, I don't like to tell people that mod_perl is "compiled" because
it's really no more compiled than any other perl script. If you want
to be precise, you could say mod_perl is a persistent daemon for
running perl code, just like servlets are a persistent daemon for
running Java.

- Perrin

Re: Why people not using mod_perl

On Thu, Sep 17, 2009 at 10:05 AM, Perrin Harkins wrote=
:
> On Wed, Sep 16, 2009 at 9:42 PM, Jenn G. wrote:
>> I think the more exact statement should be, mod_perl is compiled perl,
>> mod_perl is very fast.
>> But perl CGI...I must say it's very slow.
>
> Well, you can say CGI is slow, but Perl CGI is very fast compared to
> the alternatives. =A0Have you ever tried Java CGI? =A0Or PHP CGI? =A0They=
're
> not fast.

but nobody run Java or PHP as CGI.
the only thing I heard is somebody run php as fastcgi under lighttpd.

>
> Also, I don't like to tell people that mod_perl is "compiled" because
> it's really no more compiled than any other perl script.

mod_perl loads and compiles perl scripts only once, but CGI loads and
compiles them every time for each request.
Am I right? thanks.

Re: Why people not using mod_perl

--0-128446846-1253155633=:31217
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable

Doesn't Amazon run mod_perl/Mason? BTW, I agree with most of your poin=
ts (would debate #4,5). I may substitute the phrase "More convenient" for =
"Easier" in #3. I would also add ... #7) How many engineers are =
available to hire that know or want to work with said technology? I bu=
ilt a great platform at eBay on mod_perl/Mason that handled eBay-size traff=
ic; we ran 6 eBay sites on it. Now it is used for specialty e-commerce sol=
utions like worldofgood.ebay.com, global.ebay.com (cross-border trade), dea=
lfinder.ebay.com, etc. In fact, on the same hardware, the main eBay Java a=
pp would support ~6 threads per box; the mod_perl platform supported ~60 (p=
refork), significant CapEx and power savings (which adds up at a place like=
eBay). =0A________________________________= 0AFrom: Brad V=
an Sickle =0ATo: mod_perl list =
=0ASent: Wednesday, September 16, 2009 3:31:30 PM=0ASubject: Re: Why people=
not using mod_perl This is a mod_perl list, so I would expect =
to see Perl championed=0Apretty heavily, but Java, .net and there ilk are u=
ndoubtedly *the*=0Achoice for large web applications. I'd like to get into=
some=0Adiscussion as to why almost all *large* sites choose these language=
s. I don't have any experience developing a large application in Java,=
=0Aalthough I do have a lot of experience working on the operations side=0A=
of a large web application that is Java based. The reasons I generall=
y hear for choosing Java over mod_perl are: 1) Speed - I don't buy th=
is at all=0A2) Maintainability - I think this makes sense. Perl can be pre=
tty easy=0Ato maintain if you stick a good framework around it, but you hav=
e to=0Aseek out that framework and YOU are responsible for adhereing to it.=
=0AAll of that is inherent in Java. It also helps that Java has OO built=
=0Ain. =0A3) Easier to package and build/move code - In my experience this=
is=0Atrue. =0A4) Advantages to be gained from running on an actually appli=
cation=0Aserver - Also valid=0A5) Compatible enterprise class middleware - =
Also true, Java plugs into=0Amore truly enterprise level suff than Perl doe=
s. (security frameworks,=0Aetc... ) =0A6) Support A lot of the indus=
try seems look at Perl as obsolete technology that=0Ahas been replaced by *=
insert hot new technology of the week here* =0Awhich is a total shame. I'v=
e worked with a lot of technologies and I=0Athink Perl is a great choice fo=
r small/medium websites and webapps,=0Awhich is probably what most of us wo=
rk on. But I'm very interested to=0Aknow at what point (if any) a site/app=
grows too large or too complex=0Afor mod_perl and what defines that turnin=
g point. Could Amazon run on=0Amod_perl for example? =0A>=0A>=0A>=0A=
>>Phil Carmody wrote:=0A> =0A>--- On Thu, 9/17/09, Igor Chudov il.com> wrote:=0A>>=0A>>My site algebra.com is about 80,000=0A>>>lines of m=
od_perl code.=0A>>>=0A>>>I wrote a relatively large framework, with many ho=
megrown=0A>>>perl modules, about five years ago. =0A>>>It uses a database, =
image generation modules, a big=0A>>>mathematical engine that I wrote (that=
"shows=0A>>>work", unlike popular third party packages), etc. =0A>>>=0A>>>=
=0A>>>All pages of my site are dynamic and it is very image heavy=0A>>>due =
to math formulae. =0A>>>=0A>>>I can say two things: =0A>>>=0A>>>1) It is re=
latively fast, serving pages in 0.1 seconds or=0A>>>so=0A>>>=0A>>>2) Despit=
e the quantity of code, and its age, it is still=0A>>>very maintainable and=
understandable (to me). =0A>>>=0A>>In that case, would you like to fix its=
mangled output?=0A>>=0A>>e.g. http://www.algebra.com/algebra/homework/divi=
sibility/Prime_factorization_algorithm.wikipedia=0A>>=0A>>=C 3=82 Â (Re=
directed from Prime factorization algorithm)=0A>>=0A>>faster than O((1+=C3=
ŽÂ=B5)b) for all positive ε=0A>>=0A>>an integer M with 1 =C3=
¢â€°Â¤ M â‰Â=A4 N=0A>>=0A>>Pollard's p â=CB=
†â€™ 1 algorithm=0A>>=0A>>Section 4.5.4: Factoring into Primes, pp.=
379–417.=0A>>=0A>>Chapter 5: Exponential Factoring =
Algorithms, pp. 191–226. Chapter 6: Subexponential F=
actoring Algorithms, pp. 227–284. Section 7.4: Ellip=
tic curve method, pp. 301–313.=0A>>=0A>>Eric W. Weis=
stein, â€Å=93RSA-640 Factoredâ₠=0A>>=0A>>v =C3=
¢â‚¬Â¢ d â€Â=A2 e=0A>>=0A>>AKS · APR =
· Ballie–PSW · ECPP Â=C2=
=B7 Fermat · Lucas · Lucas–Leh=
mer · Lucas–Lehmerâ€â=80=
=9CRiesel · Proth's theorem · Pépin's =C3=
‚Â=B7 Solovay–Strassen · Millerâ=
€“Rabin · Trial division=0A>>=0A>>Sieve of Atkin=
· Sieve of Eratosthenes · Sieve of Sundaram Â=
· Wheel factorization=0A>>=0A>>CFRAC · Dixon's ·=
ECM · Euler's · Pollard's rho · P â=
ˆâ€=99 1 · P + 1 · QS · GNFS =C3=
‚Â=B7 SNFS · rational sieve · Fermat's Â=C2=
=B7 Shanks' square forms · Trial division · Shor's=0A=
>>=0A>>Ancient Egyptian multiplication · Aryabhata · =
Binary GCD · Chakravala · Euclidean · Exte=
nded Euclidean · integer relation algorithm · integer=
square root · Modular exponentiation · Schoof's =C3=
‚Â=B7 Shanks-Tonelli=0A>>=0A>>=0A>>=0A>>Looks like you've got utf8 and =
iso8859-1 messed up.=0A>>=0A>>Phil=0A>>=0A>>=0A>>=0A>>=0A>>=0A>>
--0-128446846-1253155633=:31217
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable

ad>

Re: Why people not using mod_perl

On Wed, Sep 16, 2009 at 10:15 PM, Jenn G. wrote:
> but nobody run Java or PHP as CGI.
> the only thing I heard is somebody run php as fastcgi under lighttpd.

Some cheap ISPs run PHP as CGI for security reasons. My point is, it
doesn't make any difference if Perl is slow when you run it as CGI,
since you don't need to run it as CGI any more than you need to run
Java that way.

> mod_perl loads and compiles perl scripts only once, but CGI loads and
> compiles them every time for each request.
> Am I right?

That's right. In both cases the perl code is compiled, but with
mod_perl the perl interpreter is kept in memory so it doesn't need to
recompile it each time.

- Perrin

Re: Why people not using mod_perl

--00163646be842647120473bd6dc9
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

I would also add, in addition to the frameworks, the availability of tools
such as Netbeans and Eclipse IDE's are unmatched in the perl domain. These
IDE's provide many high-level conveniences for enterprise developers, most
notably in the realm of SOA (such as graphical building of BPEL and CEP).

After nearly 10 years building and maintaining a critical government system=
,
we are sadly migrating away from mod_perl to a J2EE based solution due to
the success and growth of our mod_perl-based system. mod_perl and MySQL ha=
s
served as well when we were taking on medium-to-large loads...however, as w=
e
are growing to a distributed (multi-site, multi-node) system, with tie-ins
to numerous internal and external business systems across the enterprise,
with development partners working at distributed factories...tools such as
Netbeans and it's tight integration with Glassfish, SVN, and Hudson make
building at this level a lot more manageable. I found that mod_perl for
large-scale web applications works great, and if necessary horizontal
scaling is achievable to sustain even more load. However, when dealing wit=
h
complex SOA architectures, and the management of business workflows...the
framework support and tools to accomplish this just aren't there in perl.

Add to this Jeff's comment on the availability of high caliber perl
engineers...we are almost forced to make this decision.

We will continue to use mod_perl for other uses, such as our custom SCM/ALM
system we built over the years...but the main product is migrating.


On Wed, Sep 16, 2009 at 10:47 PM, Jeff Nokes wrote:

> Doesn't Amazon run mod_perl/Mason?
>
> BTW, I agree with most of your points (would debate #4,5). I may
> substitute the phrase "More convenient" for "Easier" in #3. I would also
> add ...
>
> #7) How many engineers are available to hire that know or want to wor=
k
> with said technology?
>
> I built a great platform at eBay on mod_perl/Mason that handled eBay-size
> traffic; we ran 6 eBay sites on it. Now it is used for specialty e-comme=
rce
> solutions like worldofgood.ebay.com, global.ebay.com (cross-border trade)=
,
> dealfinder.ebay.com, etc. In fact, on the same hardware, the main eBay
> Java app would support ~6 threads per box; the mod_perl platform supporte=
d
> ~60 (prefork), significant CapEx and power savings (which adds up at a pl=
ace
> like eBay).
>
>
>
> ------------------------------
> *From:* Brad Van Sickle
> *To:* mod_perl list
> *Sent:* Wednesday, September 16, 2009 3:31:30 PM
> *Subject:* Re: Why people not using mod_perl
>
>
>
> This is a mod_perl list, so I would expect to see Perl championed pretty
> heavily, but Java, .net and there ilk are undoubtedly *the* choice for la=
rge
> web applications. I'd like to get into some discussion as to why almost =
all
> *large* sites choose these languages.
>
> I don't have any experience developing a large application in Java,
> although I do have a lot of experience working on the operations side of =
a
> large web application that is Java based.
>
> The reasons I generally hear for choosing Java over mod_perl are:
>
> 1) Speed - I don't buy this at all
> 2) Maintainability - I think this makes sense. Perl can be pretty easy t=
o
> maintain if you stick a good framework around it, but you have to seek ou=
t
> that framework and YOU are responsible for adhereing to it. All of that =
is
> inherent in Java. It also helps that Java has OO built in.
> 3) Easier to package and build/move code - In my experience this is true.
> 4) Advantages to be gained from running on an actually application server=
-
> Also valid
> 5) Compatible enterprise class middleware - Also true, Java plugs into mo=
re
> truly enterprise level suff than Perl does. (security frameworks, etc... =
)
> 6) Support
>
> A lot of the industry seems look at Perl as obsolete technology that has
> been replaced by *insert hot new technology of the week here* which is a
> total shame. I've worked with a lot of technologies and I think Perl is =
a
> great choice for small/medium websites and webapps, which is probably wha=
t
> most of us work on. But I'm very interested to know at what point (if an=
y)
> a site/app grows too large or too complex for mod_perl and what defines t=
hat
> turning point. Could Amazon run on mod_perl for example?
>
>
>
>
>
> Phil Carmody wrote:
>
> --- On Thu, 9/17/09, Igor Chudov =
wrote:
>
> My site algebra.com is about 80,000
> lines of mod_perl code.
>
> I wrote a relatively large framework, with many homegrown
> perl modules, about five years ago.
>
> It uses a database, image generation modules, a big
> mathematical engine that I wrote (that "shows
> work", unlike popular third party packages), etc.
>
>
> All pages of my site are dynamic and it is very image heavy
>
> due to math formulae.
>
> I can say two things:
>
> 1) It is relatively fast, serving pages in 0.1 seconds or
> so
>
> 2) Despite the quantity of code, and its age, it is still
> very maintainable and understandable (to me).
>
> In that case, would you like to fix its mangled output?
>
> e.g. http://www.algebra.com/algebra/homework/divisibility/Prime_f actoriza=
tion_algorithm.wikipedia
>
> Â Â (Redirected from Prime factorization algorithm)
>
> faster than O((1+ε)b) for all positive ε
>
> an integer M with 1 â‰Â=A4 M â‰Â=A4 N
>
> Pollard's p âˆâ€=99 1 algorithm
>
> Section 4.5.4: Factoring into Primes, pp. 379–417.
>
> Chapter 5: Exponential Factoring Algorithms, pp. 191â€â=80=
=9C226. Chapter 6: Subexponential Factoring Algorithms, pp. 227ââ‚=
¬â€œ284. Section 7.4: Elliptic curve method, pp. 301ââ‚=
â€=9C313.
>
> Eric W. Weisstein, â€Å=93RSA-640 Factoredââ‚
>
> v â€Â=A2 d â€Â=A2 e
>
> AKS · APR · Ballie–PSW =C3=
‚Â=B7 ECPP · Fermat · Lucas · Lucas=C3=
¢â‚¬â€=9CLehmer ·
> Lucas–Lehmer–R iesel Â=
· Proth's theorem · Pépin's · Solovay=
–Strassen · Millerâ€â=80=
=9CRabin · Trial division
>
> Sieve of Atkin · Sieve of Eratosthenes · Sieve of S=
undaram · Wheel factorization
>
>
> CFRAC · Dixon's · ECM · Euler's Â=
· Pollard's rho · P âˆâ€=99 1 · P =
+ 1 · QS · GNFS · SNFS · ration=
al sieve · Fermat's · Shanks' square forms Â=C2=
=B7 Trial division · Shor's
>
> Ancient Egyptian multiplication · Aryabhata · Binar=
y GCD · Chakravala · Euclidean · Extended =
Euclidean · integer relation algorithm · integer squa=
re root · Modular exponentiation · Schoof's Â=C2=
=B7 Shanks-Tonelli
>
>
>
> Looks like you've got utf8 and iso8859-1 messed up.
>
> Phil
>
>
>
>
>
>

--00163646be842647120473bd6dc9
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

I would also add, in addition to the frameworks, the availability of tools =
such as Netbeans and Eclipse IDE's are unmatched in the perl domain.=C2=
=A0 These IDE's provide many high-level conveniences for enterprise dev=
elopers, most notably in the realm of SOA (such as graphical building of BP=
EL and CEP).



After nearly 10 years building and maintaining a critical government sy=
stem, we are sadly migrating away from mod_perl to a J2EE based solution du=
e to the success and growth of our mod_perl-based system.  mod_perl an=
d MySQL has served as well when we were taking on medium-to-large loads...h=
owever, as we are growing to a distributed (multi-site, multi-node) system,=
with tie-ins to numerous internal and external business systems across the=
enterprise, with development partners working at distributed factories...t=
ools such as Netbeans and it's tight integration with Glassfish, SVN, a=
nd Hudson make building at this level a lot more manageable.  I found =
that mod_perl for large-scale web applications works great, and if necessa=
ry horizontal scaling is achievable to sustain even more load.  Howeve=
r, when dealing with complex SOA architectures, and the management of busin=
ess workflows...the framework support and tools to accomplish this just are=
n't there in perl.



Add to this Jeff's comment on the availability of high caliber perl=
engineers...we are almost forced to make this decision.

We will con=
tinue to use mod_perl for other uses, such as our custom SCM/ALM system we =
built over the years...but the main product is migrating.

Re: Why people not using mod_perl

--0015175cf8a0f0cc050473c4a6c5
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Interesting. I did not even know about that #2 guy.

What sort of hardware and OS are you running there?

Igor

On Thu, Sep 17, 2009 at 12:54 AM, Jeff Nokes wrote:

> Well, actually Igor, we ended up writing eBay::API. We needed something
> that was able to extend many more web services that are internal-use only=
,
> that the public doesn't have access to. The fact that eBay web service
> data-types are probably the most complex out there, and they change often=
,
> we had to come up with a way to easily incorporate those changes by slurp=
ing
> up a giant WSDL, and auto-generating all the classes and data types, etc.
>
> But we do thank you for writing that. I knew of many API clients at the
> time that absolutely loved Net::eBay! In fact, I think at the time, the =
#2
> API client (in listings) was perl-based, and using it.
>
> Cheers,
> - Jeff
>
> ------------------------------
> *From:* Igor Chudov
> *To:* Jeff Nokes
> *Cc:* Brad Van Sickle ; mod_perl list <
> modperl@perl.apache.org>
> *Sent:* Wednesday, September 16, 2009 8:26:53 PM
>
> *Subject:* Re: Why people not using mod_perl
>
> You must have use my module Net::eBay, at some point, right?
>
> I wrote Net::eBay about 3 years ago.
>
> Igor
>
> On Wed, Sep 16, 2009 at 9:47 PM, Jeff Nokes wrote:
>
>> Doesn't Amazon run mod_perl/Mason?
>>
>> BTW, I agree with most of your points (would debate #4,5). I may
>> substitute the phrase "More convenient" for "Easier" in #3. I would als=
o
>> add ...
>>
>> #7) How many engineers are available to hire that know or want to wo=
rk
>> with said technology?
>>
>> I built a great platform at eBay on mod_perl/Mason that handled eBay-siz=
e
>> traffic; we ran 6 eBay sites on it. Now it is used for specialty e-comm=
erce
>> solutions like worldofgood.ebay.com, global.ebay.com (cross-border
>> trade), dealfinder.ebay.com, etc. In fact, on the same hardware, the
>> main eBay Java app would support ~6 threads per box; the mod_perl platfo=
rm
>> supported ~60 (prefork), significant CapEx and power savings (which adds=
up
>> at a place like eBay).
>>
>>
>>
>> ------------------------------
>> *From:* Brad Van Sickle
>> *To:* mod_perl list
>> *Sent:* Wednesday, September 16, 2009 3:31:30 PM
>> *Subject:* Re: Why people not using mod_perl
>>
>>
>>
>> This is a mod_perl list, so I would expect to see Perl championed pretty
>> heavily, but Java, .net and there ilk are undoubtedly *the* choice for l=
arge
>> web applications. I'd like to get into some discussion as to why almost=
all
>> *large* sites choose these languages.
>>
>> I don't have any experience developing a large application in Java,
>> although I do have a lot of experience working on the operations side of=
a
>> large web application that is Java based.
>>
>> The reasons I generally hear for choosing Java over mod_perl are:
>>
>> 1) Speed - I don't buy this at all
>> 2) Maintainability - I think this makes sense. Perl can be pretty easy =
to
>> maintain if you stick a good framework around it, but you have to seek o=
ut
>> that framework and YOU are responsible for adhereing to it. All of that=
is
>> inherent in Java. It also helps that Java has OO built in.
>> 3) Easier to package and build/move code - In my experience this is true=
..
>> 4) Advantages to be gained from running on an actually application serve=
r
>> - Also valid
>> 5) Compatible enterprise class middleware - Also true, Java plugs into
>> more truly enterprise level suff than Perl does. (security frameworks,
>> etc... )
>> 6) Support
>>
>> A lot of the industry seems look at Perl as obsolete technology that has
>> been replaced by *insert hot new technology of the week here* which is =
a
>> total shame. I've worked with a lot of technologies and I think Perl is=
a
>> great choice for small/medium websites and webapps, which is probably wh=
at
>> most of us work on. But I'm very interested to know at what point (if a=
ny)
>> a site/app grows too large or too complex for mod_perl and what defines =
that
>> turning point. Could Amazon run on mod_perl for example?
>>
>>
>>
>>
>>
>> Phil Carmody wrote:
>>
>> --- On Thu, 9/17/09, Igor Chudov =
wrote:
>>
>> My site algebra.com is about 80,000
>> lines of mod_perl code.
>>
>> I wrote a relatively large framework, with many homegrown
>> perl modules, about five years ago.
>>
>> It uses a database, image generation modules, a big
>> mathematical engine that I wrote (that "shows
>> work", unlike popular third party packages), etc.
>>
>>
>> All pages of my site are dynamic and it is very image heavy
>>
>> due to math formulae.
>>
>> I can say two things:
>>
>> 1) It is relatively fast, serving pages in 0.1 seconds or
>> so
>>
>> 2) Despite the quantity of code, and its age, it is still
>> very maintainable and understandable (to me).
>>
>> In that case, would you like to fix its mangled output?
>>
>> e.g. http://www.algebra.com/algebra/homework/divisibility/Prime_f actoriz=
ation_algorithm.wikipedia
>>
>> Â Â (Redirected from Prime factorization algorithm)
>>
>> faster than O((1+ε)b) for all positive ε
>>
>> an integer M with 1 â‰Â=A4 M â‰Â=A4 N
>>
>> Pollard's p âˆâ€=99 1 algorithm
>>
>> Section 4.5.4: Factoring into Primes, pp. 379–417=
..
>>
>> Chapter 5: Exponential Factoring Algorithms, pp. 191â€â=
€œ226. Chapter 6: Subexponential Factoring Algorithms, pp. 227â=E2=
‚¬â€=9C284. Section 7.4: Elliptic curve method, pp. 301ââ‚=
¬â€œ313.
>>
>> Eric W. Weisstein, â€Å=93RSA-640 Factoredââ‚
>>
>> v â€Â=A2 d â€Â=A2 e
>>
>> AKS · APR · Ballie–PSW =C3=
‚Â=B7 ECPP · Fermat · Lucas · Lucas=C3=
¢â‚¬â€=9CLehmer ·
>> Lucas–Lehmer–R iesel =C3=
‚Â=B7 Proth's theorem · Pépin's · Solo=
vay–Strassen · Millerâ€â=
€œRabin · Trial division
>>
>> Sieve of Atkin · Sieve of Eratosthenes · Sieve of =
Sundaram · Wheel factorization
>>
>>
>> CFRAC · Dixon's · ECM · Euler's Â=
· Pollard's rho · P âˆâ€=99 1 · P =
+ 1 · QS · GNFS · SNFS · ration=
al sieve · Fermat's · Shanks' square forms Â=C2=
=B7 Trial division · Shor's
>>
>> Ancient Egyptian multiplication · Aryabhata · Bina=
ry GCD · Chakravala · Euclidean · Extended=
Euclidean · integer relation algorithm · integer squ=
are root · Modular exponentiation · Schoof's Â=
· Shanks-Tonelli
>>
>>
>>
>> Looks like you've got utf8 and iso8859-1 messed up.
>>
>> Phil
>>
>>
>>
>>
>>
>>
>

--0015175cf8a0f0cc050473c4a6c5
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Interesting. I did not even know about that #2 guy.

What sort of har=
dware and OS are you running there?

Igor