python is not a functional language, but it does offer lambda, filter, reduce, list comprehensions, and python's functions are first class objects. there is also a good chance that it will offer continuations in a future version as well. python can also construct and execute code on the fly, as well as offering support for "deep lexicals" at the syntactic level.
as far as metaprogramming goes, it offers better support than OO language i know of that has a modern grammer. http://www.python.org/doc/essays/metaclasses/
python is great mix of OO with a dash of functional programming and metaprogramming power.
who wants macros. have you ever worked with MFC - gag!
the only part of python that is similar to perl is the excellent regexp library. and since perl got it right, why not adopt it?
perl is a heavy investment. if you learned perl you should stick with it. perl is open and very well supported.
but if...
- you feel trapped with a proprietary or single platform language.
- you like java's OO and libraries but miss the syntactic expressiveness of functional languages or meta programming.
- you like vb's ease of use (try pythonwin by mark hammond - code completion, integrated interpreter window for tweaking variables and expressions on the fly)
- you like opengl/3D (pyopengl, blender, and truespace all use python)
- you think having one draw api and multiple backends is cool (such as illustrator, pdf, postscript, opengl, various widgets,...) see piddle project on sourceforge
- you need software engineering constructs in a rapid application package like: asserts, easy testing at package level (all packages are runnable externally or in a self test mode), full introspection, built in documentation that is also available via introspection, the debugging power of being able to intercept any method call before it happens, or the rapid feedback of transparent bytecode compliation.
then, maybe python is for you.
what you will NOT find in python (that i know of):
- support for a transactional container like mts / EJB. there are libraries for talking to COM and CORBA. you will have to rely on a database or bridge to another system for transactional applications at the object level.
- blazing speed. you can create python objects in 'c' for this.
retake a comp sci 101 refresher. most languages will not open a file without a full path to the file. python's file operations are just a thin wrapper around ansi c's file operations.
it would incorrect for "open" to make assumptions about where the file is.
excerpt from http://www.electronicnews.com/issue/RegisteredIssu es/2000/05152000/z54f-1.asp
Cost effectiveness. Cost is very difficult to compare, because a number of variables can influence the final cost of the product. Another I/O standard, 1394 (FireWire), achieves similar speeds to USB 2.0 and thus is a good model for comparison. The most widely accepted comparison standard is the use of gate counts to provide an apples-to-apples comparison. Vendors who supply USB 2.0 and 1394 solutions estimate that 1394 host controllers have two to three times more gates than do USB 2.0 controllers. Vendors also estimate that 1394 peripheral controllers have four to five times more gates than comparable USB 2.0 peripheral silicon. With Intel planning to integrate USB 2.0 host controllers into future chipsets, the system costs of implementing USB 2.0 will be incremental vs. the cost of adding a discrete host controller for something like 1394.
3rd paragraph, second page... (http://www2.discreet.com/events/siggraph2000/pdf/ gmax.pdf)
"When the first set of supported game packs is released, Discreet plans to have no charge downloads of 3D Studio gMax available to consumers."
You can also gleen, that while Discreet is giving gMax away, they will be licensing and distributing Game Packs (the set of specific tools for extending a specific game?). I wonder how the mostly free mod community feels about this?
my understanding was that gnome from the start intended to bring an object model to linux. why would the gnome team spend time working on porting other non-componentized applications? in order for the component to be useful, the underlying object model has to be exposed as well. essentially that's a rewrite if the former code was not written with that in mind.
i can't speak for miguel, but i doubt highly that he wants the entire microsoft framework for developing applications on linux. maybe at a COM library and up level, but i'm sure he wouldn't wish MFC or the registry on anyone. (MFC is dying out by the way, and i haven't heard the term vbx used since 96, you really aren't up to speed on nt development. and, if you are going to complain about the win32 api, to be fair you will have to lump bare metal calls in X as well, since the win32 includes the GDI.)
regarding your software engineering talents... we agree to disagree;>
miguel works in the (graphical) application framework area of linux - how applications are built, the resources they share, the services they expect from a window manager.
since there is no 'standard' for the application services that window managers provide, new end user applications often have to each rewrite this code because they cannot count on it being provided in a logical, consistent manner.
the problem is not in what miguel is saying. it is that you don't understand the context.
without a this framework in place the effort required to create linux applications is sub-optimal.
jim
ps. you really have no idea what you are talking about with DOS, NT, or software engineering.
Bad programmers can create unmaintainable code using any tool. By not quantifying what makes your other projects maintainable you provide a high noise to signal post. In the meanime, I would suggest you add Design Patterns by Gamma, et al to your reading list for understanding OO architecture and design from an engineering aspect if you haven't done so. An alternative OO solution to the Zope/Python duo would be the Java/Enhydra solution (it's next on my list to evaluate.)
I'm just starting out on zope, so i don't have too much to say. But there are a few things that should preclude you from using zope:
1) If you don't understand OO architectures. Zope is OO from the ground up, and because documentation is sparse, you will need to "get" OO in order to use it, specifically, Digital Creations use of OO.
2) If you have no web experience. Don't start with Zope. Zope leverages an internal DOM, a scripting language (DTML), user defined objects (Zclasses), a built in OO database, and python methods to get the job done. Clearly not for beginners.
3) Static sites. Zope is highly dynamic and would probably be overkill.
4) High volume combined with intensive server side processing without a distributed architecture. Zope is built with Python. Python is interpreted byte code. If you have intensive processing going on and you want the benefits of a highly dynamic environment, you will pay for it in CPU and Memory use.
Why you might use Zope: 1) You are an OO developer. Zope supports containers and user objects. In Zope, everything is an object with properties. Zope supports inheritence through its folder containers, thus allowing child folders to benefit from parent properties. If used in this regard you can build some very easy to maintain sites. Zope has built in persistence and indexing available via inheritence. Even the concept of users and groups are built in, and attachable to any object in the system.
2) You need a mid size, highly dynamic website with an emphasis on content management. this is where Zope excels IMHO.
3) You need to build server side web applications that will run on NT or Linux or BSD...
i said it wasn't free or open. no arguement here:> (BUT, check out boa, potentially a better open source alternative, because it is also cross platform. http://boa-constructor.sourceforge.net/ )
>>2) > as a plus, they use tcl/tk How is the use >>of TCL in a supposed Python tool a plus?
multiplatform, baby! they also have an abstract layer, so they can support other widget toolkits in the future. just say no to platform lock!
(more python exposure is great, but taking a multi-platform language and tying it down to one platform isn't.)
pythonware has had a "visual" python builder for many months. as a plus, they use tcl/tk + their advanced python imaging library for gui, so their solution is truly multi-platform. it's not free or open, but it is available now (their web site FAQ says they will make an evalution version available july 20th.)
www.brooktrout.com these guys have just ported many of their IP telephony products to Linux. http://www.brooktrout.com/pages/news/press/2000/06 27linuxnewnetwork.html jim
Slashdot Mirror
first off - moderator, kiss my informative ass ;>
python is not a functional language, but it does offer lambda, filter, reduce, list comprehensions, and python's functions are first class objects. there is also a good chance that it will offer continuations in a future version as well. python can also construct and execute code on the fly, as well as offering support for "deep lexicals" at the syntactic level.
as far as metaprogramming goes, it offers better support than OO language i know of that has a modern grammer. http://www.python.org/doc/essays/metaclasses/
python is great mix of OO with a dash of functional programming and metaprogramming power.
who wants macros. have you ever worked with MFC - gag!
the only part of python that is similar to perl is the excellent regexp library. and since perl got it right, why not adopt it?
jim
perl is a heavy investment. if you learned perl you should stick with it. perl is open and very well supported.
...) see piddle project on sourceforge
but if...
- you feel trapped with a proprietary or single platform language.
- you like java's OO and libraries but miss the syntactic expressiveness of functional languages or meta programming.
- you like vb's ease of use (try pythonwin by mark hammond - code completion, integrated interpreter window for tweaking variables and expressions on the fly)
- you like opengl/3D (pyopengl, blender, and truespace all use python)
- you think having one draw api and multiple backends is cool (such as illustrator, pdf, postscript, opengl, various widgets,
- you need software engineering constructs in a rapid application package like: asserts, easy testing at package level (all packages are runnable externally or in a self test mode), full introspection, built in documentation that is also available via introspection, the debugging power of being able to intercept any method call before it happens, or the rapid feedback of transparent bytecode compliation.
then, maybe python is for you.
what you will NOT find in python (that i know of):
- support for a transactional container like mts / EJB. there are libraries for talking to COM and CORBA. you will have to rely on a database or bridge to another system for transactional applications at the object level.
- blazing speed. you can create python objects in 'c' for this.
wishing i had a paying python job,
jim
retake a comp sci 101 refresher. most languages will not open a file without a full path to the file. python's file operations are just a thin wrapper around ansi c's file operations.
it would incorrect for "open" to make assumptions about where the file is.
jim
600-700mhz PIII
128mb ram
40G HD
nvdia 3d
built in ethernet
all it needs is a voice activated mic and you've got a mini-holodeck.
jim
terrible post. slashdot trend since ipo: increased quantity, reduced quality. jim
excerpt from http://www.electronicnews.com/issue/RegisteredIssu es/2000/05152000/z54f-1.asp
Cost effectiveness. Cost is very difficult to compare, because a number of variables can influence the final cost of the product. Another I/O standard, 1394 (FireWire), achieves similar speeds to USB 2.0 and thus is a good model for comparison. The most widely accepted comparison standard is the use of gate counts to provide an apples-to-apples comparison. Vendors who supply USB 2.0 and 1394 solutions estimate that 1394 host controllers have two to three times more gates than do USB 2.0 controllers. Vendors also estimate that 1394 peripheral controllers have four to five times more gates than comparable USB 2.0 peripheral silicon. With Intel planning to integrate USB 2.0 host controllers into future chipsets, the system costs of implementing USB 2.0 will be incremental vs. the cost of adding a discrete host controller for something like 1394.
cheers,
jim
USB uses a microcontroller architecture, not a digital signal processor.
http://www.cypress.com/usb/fx2factsheet.pdf
also, intel sold their usb chip division some time ago.
jim
3rd paragraph, second page.../ gmax.pdf)
(http://www2.discreet.com/events/siggraph2000/pdf
"When the first set of supported game packs is released, Discreet plans to have no charge downloads of 3D Studio gMax available to consumers."
You can also gleen, that while Discreet is giving gMax away, they will be licensing and distributing Game Packs (the set of specific tools for extending a specific game?). I wonder how the mostly free mod community feels about this?
jim
i can't speak for miguel, but i doubt highly that he wants the entire microsoft framework for developing applications on linux. maybe at a COM library and up level, but i'm sure he wouldn't wish MFC or the registry on anyone. (MFC is dying out by the way, and i haven't heard the term vbx used since 96, you really aren't up to speed on nt development. and, if you are going to complain about the win32 api, to be fair you will have to lump bare metal calls in X as well, since the win32 includes the GDI.)
regarding your software engineering talents... we agree to disagree ;>
jim
miguel works in the (graphical) application framework area of linux - how applications are built, the resources they share, the services they expect from a window manager.
since there is no 'standard' for the application services that window managers provide, new end user applications often have to each rewrite this code because they cannot count on it being provided in a logical, consistent manner.
the problem is not in what miguel is saying. it is that you don't understand the context.
without a this framework in place the effort required to create linux applications is sub-optimal.
jim
ps. you really have no idea what you are talking about with DOS, NT, or software engineering.
I'm just starting out on zope, so i don't have too much to say. But there are a few things that should preclude you from using zope:
1) If you don't understand OO architectures. Zope is OO from the ground up, and because documentation is sparse, you will need to "get" OO in order to use it, specifically, Digital Creations use of OO.
2) If you have no web experience. Don't start with Zope. Zope leverages an internal DOM, a scripting language (DTML), user defined objects (Zclasses), a built in OO database, and python methods to get the job done. Clearly not for beginners.
3) Static sites. Zope is highly dynamic and would probably be overkill.
4) High volume combined with intensive server side processing without a distributed architecture. Zope is built with Python. Python is interpreted byte code. If you have intensive processing going on and you want the benefits of a highly dynamic environment, you will pay for it in CPU and Memory use.
Why you might use Zope: 1) You are an OO developer. Zope supports containers and user objects. In Zope, everything is an object with properties. Zope supports inheritence through its folder containers, thus allowing child folders to benefit from parent properties. If used in this regard you can build some very easy to maintain sites. Zope has built in persistence and indexing available via inheritence. Even the concept of users and groups are built in, and attachable to any object in the system.
2) You need a mid size, highly dynamic website with an emphasis on content management. this is where Zope excels IMHO.
3) You need to build server side web applications that will run on NT or Linux or BSD ...
4) You need an excuse to learn Python ;>
Cheers,
Jim
i said it wasn't free or open. no arguement here :> (BUT, check out boa, potentially a better open source alternative, because it is also cross platform. http://boa-constructor.sourceforge.net/ )
>>2) > as a plus, they use tcl/tk How is the use >>of TCL in a supposed Python tool a plus?
multiplatform, baby! they also have an abstract layer, so they can support other widget toolkits in the future. just say no to platform lock!
jim
pythonware has had a "visual" python builder for many months. as a plus, they use tcl/tk + their advanced python imaging library for gui, so their solution is truly multi-platform. it's not free or open, but it is available now (their web site FAQ says they will make an evalution version available july 20th.)
http://www.pythonware.com/products/works/faq.htm
jim
www.brooktrout.com these guys have just ported many of their IP telephony products to Linux. http://www.brooktrout.com/pages/news/press/2000/06 27linuxnewnetwork.html jim