We want to write a Linux service in php and compile it with HIPHOP. Since we started the project with php and we could do all the programming in-house instead of hiring a c++ programmer etc. we would love to stick to php. Speed in execution is not (so) relevant for us since the daemon is just doing some monitoring but we would like to close up the code to obfuscate it. The daemon will do some network communication and logging to a db. Is this a viable route to go? In another post someone described that hiphop needs special attention in programming since not all php features are implemented. Is this still the case? I would love to here your overall opinion on our idea.
HIPHOP is quite a beast to handle. It is very limited, so it depends specifically on your application and where it will be deployed. Remember, at present it only runs on 64 bit architectures .. so if you wanted to deploy on a 32-bit machine, you are immediately stuck in the mud.
You may have to build many different binaries for different linux distro's depending on the nature of your application. Since HipHop only works well on Fedora and CentOS, you are severely limiting your scope. Once you move off of the PHP interpreter, you lose a very large amount of interchangeability between operating systems (Think about it: Windows, Virtually all Linux, All Major BSD Distributions, ... )
Also keep in mind, I'm not sure to what extent you want to "obfuscate" your code. If you want to make network calls, etc and keep those hidden as well, a packet sniffer can see exactly how you are communicating with the outside world extremely easily.
Likewise, a debugger and a reasonably seasoned programmer will be able to reverse engineer your binary to a larger degree than you may be aware.
You may want to look into alternatives such as Zend Encoder or IonCube Encoder would be the preferred method to go about things, but these are non-free options. There are other encoders out there as well that you may want to look into.
I'm not exactly sure what you're doing other than "monitoring", so I can't say for sure. But a secondary option would be simply to severely limit the amount of code that is being run on the client machines (assuming they are reporting to a server machine) and let the server machines, which are assumed in your total control, handle more processing if any way possible.
I invite you to simply explore the idea yourself by testing, since once again, it's extremely dependent on the nature of your application and where you intend to deploy it. (And for many people, something like "where to deploy" can change rapidly). HipHop was created with a very narrow scope: Run PHP code as fast as possible. It isn't designed to be highly flexible or highly interchangeable between OS's and CPU Architectures. Please consider this before you write a large application reliant on it, and please make sure you fully understand every implication of using HipHop. Test, test, test.
I'm looking to develop a Win32 desktop app - a one off, for a personal need. A GUI is not scrictly needed, though would be a bonus.
What is needed:
The ability to monitor the window title of a specific window.
The ability to run DOS commands when this window title changes.
I hope my programming ability is up to this - I feel I'm pretty good with PHP, but I'm not ready to spend time learning OO for this one-off project.
What non OO (or at least not mandatorily OO) languages are there for desktop apps that might be suitable for a beginner on a task like?
Any other, more hacky approaches are welcome too - batch scripts etc.
Many thanks for any advice,
Jack
Whether or not the language supports OO doesn't really matter for your purposes. After all, PHP supports objects and you seem to do just fine with it.
Personally, i'd recommend Java or C# to get started with. The communities for these two languages are huge and there are plenty of tutorials online to help you get started.
It's extremely easy to get starting writing C# with Visual Studio Express. And a good hello world tutorial.
Also, if you stick with C# you can take advantage of WMI which will allow you to do everything you need for this project (and much much more).
Lastly, most windows machines will be able to run your application without having to install anything extra and Visual Studio builds the .exe for you as part of the build process.
You can use PHP for desktop apps if you really want to. Just install the php CLI.
You can even do a gui for your desktop app in php: http://gtk.php.net/
EDIT: I'm not sure how easily you can call win32 api functions from PHP, however. There look to be a few articles about this online and a SO question: How to call winapi functions from PHP?
I would vote for Python using the included TkInter module for GUI. Dead simple to use.
Widgets aren't the prettiest looking, but development is rapid.
EDIT: I mistook "non-OO" in the question for "OO". Python is most definitely not "non-OO", but but is very well suited to doing what you asked.
You could write this in pretty much any mainstream language supported by Windows. C or C++ are obvious choices. C# and Visual Basic .NET are going to require the .NET Framework ... not a bad thing, but perhaps more than you want to tackle for a simple project. Come to think of it, you might be able to do this with JScript or VBScript, although I'm not clear on what API functions you have easy access to. And I have to believe that it's possible to do with PowerShell with just a little work.
Your options are wide open.
F#
It's an awesome piece of work, has access to the Framework class libraries, supports GUI development, really easy parallel programming, compiles to IL (same as C#) but has a really concise functional syntax.
OK, go easy on me as it's my first question here. ;)
I am experienced in Web Application development, but I need to design 2 different desktop applications at work. I need advice on what languages to use (I use PHP for Web Applications, I have little other experience), any frameworks that may help me accomplish this easier and any IDEs to use to make my life easier too.
My first application will essentially be an interface to an embedded or remote DB and essentially it needs to do not much more than CRUD. A fulltext search function will be necessary as will sorting of the CRUD table too. I'd like to include the possibility of hiding columns of data too, but this isn't a priority. I currently use MySQL, but am happy to use any other DB, as long as they support importing from Excel. Of course, if it supports an excel spreadsheet as the database, fantastic, but wishful thinking...
I only need to support Windows XP upwards, this won't change. Let me know if I can add anything to help answer this.
My second application is going to be more complicated, I guess. It needs to output the SHA, MD5, CRC-32, CRC-16 signatures for an entire external drive on-screen and/or into a text document of the users choice (new and append). The external drive could be Compact Flash, EEPROM, CD-ROM, DVD-ROM, SD, etc. Whereas I have an idea where to start on the first application, this one is where I need major help as I have no idea which programming language has built-in support for all the algorithms above, or the best way to go about coding it. So any and all advice will be more than welcome.
Edit: OK, so I have come across Titanium and am wondering if this is the best option out there for my 2 projects. The first one I am certain I can achieve, but the second project is going to be pretty difficult I believe, as unlike most hash calculators out there, I do not only need to check individual files for SHA/MD5/CRC-32/CRC-16, but I also need to check Compact Flash through a USB-connected Sandisk CF reader and EEPROMs through a COM or USB connection as well. I essentially need to produce the signature/hash for the entire drive, not the files on it. Any advice on this in particular would be helpful, please let me know if I am going too far off-topic by stackoverflow standards, as I can create a new question with this in.
I'm going to have to say Perl might be a good (though not necessarily the best) option for you, if you aren't one of the PHP developers who are Perl-allergic:
Perl is probably easier for a PHP developer to pick up than some statically typed compiled language that typical desktop apps are written in
Perl has libraries to interface to pretty much ANY database back-end you want, though I'm not 100% sure about using Excel as DB directly (I'm fairly confident something exists in CPAN but can't recommend anything from personal experience).
Perl has Windows GUI development - from an easier (for you) web GUI using local/embedded web server to Tk to Win32 GUI modules (StackOverflow has a couple of questions with very good references on the topic of Perl Windows GUI options).
Perl has a fairly complete set of support for MD5/SHA etc... Again CPAN is your friend.
If the only development experience you have is PHP for web, your options for quick desktop development are seriously limited.
since i don't know of any (sane) way to use PHP for desktop apps, my first thought was to reach for a framework that would let you use JavaScript and a mostly declarative UI. That would save you from most of the pain.
The best one for that would be Qt, which is supposed to let you do almost everything on QML and JavaScript. the unfortunate part is the 'almost'; since anything that escapes that would mean C++ development. The best C++ environment, for sure; but still widely different from Web work (specially PHP web work).
The other option, which I think would be best for you is Adobe AIR. I haven't really checked it; but it lets you write a 'desktop-like' app using only Flash (with or without the Flex framework) or HTML+JavaScript. When i first saw it seemed pretty much complete. For DB work i think it can use an embedded SQLite library; or access any HTTP service (which could be a local DB+PHP app exposing the CRUD methods)
You could try PHP-GTK.
It may not be the fastest, but it does work.
Is the LAMP (Linux, Apache, MySQL, PHP / Ruby / Python) stack appropriate for Enterprise use?
To be clear, by "Enterprise", I mean a large or very large company, where security, robustness, availability of skill sets, Total Cost of Ownership (TCO), scalability, and availability of tools are key considerations. Said another way, a company that looks for external adoption of frameworks / architecture - Something ubiquitous will be seen as more "valid" than something exotic / esoteric in this kind of environment.
I've seen use cases where Oracle, IBM, and Sun have implemented systems on the LAMP stack for various Enterprises. I've also seen examples where websites like yellowpages.com (Ruby on rails) and Facebook (php) are built on it. However, none of these examples are exactly what I'm looking for.
I'm really trying to find examples where it is an Enterprise standard at a very large bank (I.e., Citigroup), Telecom company (I.e., AT&T), or manufacturer (I.e., Proctor and Gamble). Just to be clear, I'm not looking for an example where it's used in a limited sense (Like at JPMorgan Chase), but where it's a core platform for systems like CRM, manufacturing systems, or HR management, as well as for internal and external websites.
The perception I've seen so far is that applications built on the LAMP stack perform slower and are less flexible. Some of the arguments I've heard are:
Linux is seen as not as well supported as Unix, Solaris, or Windows Servers.
Apache is harder to configure and maintain than web servers like BEA WebLogic or IIS.
MySQL is a "not ready for prime time" DB for hobbyists, and not a competitor for SQL Server or Oracle (Although PostgreSQL seems to have a reputation for being more robust).
PHP / Ruby on rails are optimized for CRUD (Create, Read, Update and Delete operations). Although this is an advantage when building CRUD-intensive web aplications, both perform slower than Java/Java EE or C# (which are both common Enterprise standards). Furthermore, a lot of applications and systems (like manufacturing systems) have a lot of non-CRUD functionality that may be harder to build with PHP or Ruby, or even Python.
Can anyone please provide arguments to support or refute the idea of the LAMP stack being appropriate for the Enterprise?
Thanks!
KA
UPDATE: Some times the LAMP Stack is Appropriate for Enterprise Use: Externally-Facing Blogs
"but where it's a core platform for systems like CRM and HR, as well as for internal and external websites"
First, find a LAMP CRM or HR application.
Then find a customer for the LAMP CRM or HR application.
Sadly, there aren't a lot of examples of item 1. Therefore, your case is proven. It can't be used for enterprise applications because -- currently -- there aren't any of the applications you call "enterprise".
Your other points, however, are very interesting.
Linux is seen as not as well supported as Unix, Solaris, or Windows Servers. I think Red Hat would object strongly to this. Give them a call. I think they'll make a very persuasive sales pitch. Read their success stories.
Apache is harder to configure and maintain than web servers like BEA WebLogic or IIS. By whom? Apache web site managers? Or IIS web site managers? This is entirely subjective.
MySQL is a "not ready for prime time" DB. Take it up with Sun Microsystems. I think they'd object strongly to this. Give them a call. I think they'll make a very persuasive sales pitch. Read their success stories.
PHP / Ruby on rails are optimized for CRUD, and both are slowly performing. Could be true. Java and Python might be faster. PHP and Ruby aren't the last word in LAMP.
Something ubiquitous will be seen as more "valid" than something exotic / esoteric in this kind of environment.
Although I personally wouldn't recommend PHP due to the many flaws in the language, it's most certainly ubiquitous. With the advent of phusion passenger, Rails support amongst shared-hosting companies is growing pretty quickly too. I give it another year or 2 at most before 90+% of shared-hosting accounts support rails out of the box. If that's not ubiquitous, what is?
Linux is seen as not as well supported as Unix, Solaris, or Windows Servers.
If this bothers you, purchase support from RedHat, or install Solaris and purchase support from Sun. Both of those will give you just as good support as Microsoft is likely to
Apache is harder to configure and maintain than web servers like BEA WebLogic or IIS.
I can't speak for BEA WebLogic, but having configured both Apache, IIS, and Tomcat, Apache is the easiest both to understand, and to find examples and documentation for by a long way.
MySQL is a "not ready for prime time" DB for hobbyists, and not a competitor for SQL Server or Oracle.
Oh really?. You should make it your mission to tell NASA, Google, CERN, Reuters etc that they're all using a hobbyist database that isn't ready for prime-time.
PHP / Ruby on rails are optimized for CRUD, and both perform slower than Java/Java EE or C# (which are both common Enterprise standards).
There are 2 things here:
Optimized for CRUD - This is totally irrelevant.
Rails and some of the python/php frameworks are optimized for CRUD apps. Many of the C#/Java frameworks are also optimized for CRUD apps. However, if the app you're building is a CRUD app (and 99% of web applications are), isn't this a Good Thing?
If you're not building a CRUD app, there are plenty of non-crud-optimized frameworks in ruby/python/php/java/C#. Net win: Nobody (hence it's irrelevant)
Perform slower than Java/C# - This is undoubtedly true, but it also doesn't matter. For a low-traffic site the performance difference isn't going to amount to anything, and for a high-traffic site your bottleneck will be the database, whether it be MySQL, oracle, or whatever.
What you trade-off for all of this is development time.
Once you've used all this advice to convince your boss that you won't lose out on anything by using LAMP, If you crunch the numbers and show your them that it is going to take 6 man-months to build the site in Java, and only 3 to build it in ruby/python then that's really what it comes down to.
If you hire idiots to implement it, C++ & Oracle will fail to scale.
If you hire people who are smart and get things done, PHP & MySQL will scale just fine.
Same argument goes for security & robustness.
Facebook, Digg, portions of Yahoo run on PHP.
Of course, they hire lots of PhD programmers.
Just thought I'd add another website to the list of those that run on LAMP - Wikipedia. Seventh biggest website in the world, written entirely in PHP and runs off MySQL, and they only have two or three paid developers. Of course, they have some assistance from volunteers, but it's not a lot, and it's scaled just fine. Don't know if you'd really call them 'enterprise', but for such a huge and popular website they seem to have done alright for themselves.
Linux is seen as not as well supported
as Unix, Solaris, or Windows Servers.
As others have said above, give Red Hat a call and I'm sure they'll beg to differ. And the amount of support out there for Linux absolutely free is astonishing.
Apache is harder to configure and
maintain than web servers like BEA
WebLogic or IIS.
That depends who you're asking. People who usually administer IIS servers will probably view it this way. People who usually administer Apache won't. It depends on who you hire, and if your stack is LAMP you won't want to be hiring people with no Apache experience anyway.
I just want to add that I've witnessed many times that clients only feel comfortable once they dish out serious $$$ for some solution, even if it makes enterprise integration even harder, despite what arguments you bring to the table.
I think the first criteria should be your team's skill level, comfort level jut to make sure what ever platform decisions are made works well with them. Whatever you decide think of scalability and maintainability of your code. Tools are awesome no matter what stack you choose.
I personally would break it down into 3 stacks-
The Java Stack where you have Solaris or Enterprise Linux like ( RedHat ) with Weblogic/Websphere/Tomcat etc and Java Enterprise along with Hibernate,Spring etc technologies. Most would opt for Oracle as DB.
The Microsoft Stack with some Open Source if needed Win Server - IIS - .net/C# (ASP.net etc) - NHibernate, NUnit (unit testing) etc. Most likely you would want to use SQL Server as DB
None of the above stack with Enterprise Linux running a whole buffet of open source stuff like MySQL (now under Sun's domain so can be looked at seriously), Apache (there are apache gurus out there), Ruby ( not my personal choice)/ PHP (good luck) / Python (I like it because its a mature language). I would advocate python or ruby from the managing code point of view. Maybe for some it could be PHP..i am not into it.
strictly a subjective opinion but I personally find MySQL and to a lesser extent PHP to be a bit of a weakness, but certainly there's plenty of people who disagree and big companies who went LAMP.
I'd prefer to see postgres or even SQLite take chunks out of the MySQL market, and I'd like to see mono or jsp or cocoon based apps more. I guess LAMP is a bit too specific for an umbrella term. :)
Linux/Apache are hardened, lean and each comes with plenty of people(for the right price of course) who will provide support, plenty of useful tools, many at exceptionally high levels of utility which work with them and which have been built upon them.
Not sure about the other two, however. In particular MySQL seems to have taken a strange turn for the worse since their being acquired by Sun, contrary to the posts in this thread suggesting that Sun may be a good influence:
http://www.reddit.com/r/programming/comments/7gb8j/oops_we_did_it_again_mysql_51_released_as_ga_with/
The reason for not finding Enterprise applications built on LAMP is not because they aren't enterprise level but something entirely different in my opinion. A lot of the big players use LAMP or similar--Facebook and MySpace immediately come to mind. So its clearly not an issue of scale and perf.
That said, the reason I find that there aren't any enterprise apps built on LAMP is because of their intrinsic open nature. I don't want to build an actuarial module as a PHP file because anyone can steal the logic. On the other hand if I have a DLL I can retain control. You don't find a lot of 30-trial apps built on PHP for this very reason but it's much easier to achieve that kind of protection with say ASP.NET.
You have some real bad myths in your posting:
JavaEE Myths:
-App Servers easier to configure than apache, nope apache is easier.
-You imply that only JavaEE full solution is enterprise, nope.
CRUD Myths:
-CRUD is slower than JavaEE? WTF? POJO and EJB is using CRUD.
The limiting factor is not crud, its server throughput
There are 3 limiting bottleneck areas no matter what technology even MS..server implementation, persistence layer, and app layer..the technology chosen is not the speed factor as you can exchange advantages in one layer for disadvantages in the other layer.
Fro example we could spee dup Java by using document store instead of normal DB..
Most new Rails implementations use non apache servers that are faster by a factor of 3 to 5 than Apache..even a well tunned Apache server can outperform some javaEE stacks..just ask yahoo as they use Symfony on some of their properties..
I think you will find that many enterprises use Linux servers, often supported by Redhat, Novell or IBM, and that Apache is also commonly used.
But many enterprises tend to use databases like Oracle or IBM DB2 instead of open source offerings - although there are many enterprises that don't really need the kind of power those systems provide and could get away with MySQL or PostgreSQL.
And for the web-server language, I think you can use just about anything. However, if you use Apache it is probably easier to use PHP, Ruby or Python, whereas if you use IIS or Weblogic or Domino it will be easier to do it in Java / C#.
IMO there are no good general arguments against Linux and Apache; You can certainly get enterprise-level support for Linux if you're prepared to pay for it (and a good approximation of it for free if you're willing to play by the community's rules). And Apache is not that hard to configure unless you need its more complex features, which is unlikely in an application server.
You can certainly make a case against MySQL since some of the most important features in regard to data safety have been added only recently. If you're concerned about that, use PostgreSQL instead.
As for the language you write your app in: PHP has definitely proven to be able to run extremely large and complex systems; I'd be more concerned about maintainability than performance. And Ruby on Rails is "optimized for CRUD" only in asmuch as a simple CRUD webapp can be written in nearly no time (literally minutes), but that does not mean it is somehow less suited to more complex apps, just that it will take much more time (still less than with many other languages)
I suppose that large commercial CRM and HR applications might be biased toward delivering large commercial RDBMS products as the foundation for their products. If nothing else they will I'm sure prefer to unite against a common threat.
And they have a harder time justifying license and support fees if they integrate products that don't have them.
My 2c:
Linux: Since kernel 2.6 came out, I would say it is definitelly a high-quality OS. Version 2.4 wasn't quite there and 2.2 was a joke, but 2.6 is really good. Be careful with a choice of distribution, though. In my experience, RedHat/CentOS is very good, and apparently Debian (original, not Ubuntu!) can be set up nicely if you have a good admin. My experience with OpenSUSE was not very good.
Apache: Haven't used it, but I don't see why it would be a problem.
MySQL: This is the weakest point of the stack. I am not going to go into details here - look into comments at reddit.programming if you are interested. Better look at PostgreSQL.
PHP/Perl/Ruby/Python: I have worked with Perl and to a lesser extent with Python. They are probably OK for web-based applications where the bulk of the work is done by the web server and DBMS anyway. However, I do prefer static type system and would rather pick Java/C# for a business application and C++ for system programming.
I would like to suggest that we identify the scalability requirements of Enterprise systems and how they differ compared to Web Applications. Look at some of the most scalable systems like Wikipedia, Flickr, Wordpress, Facebook, MySpace and a host of others. You will see LAMP stack there. I am more of a Python fan (since I feel that the language has a cleaner feel) but I listen to experts like Cal Henderson (Flickr) who wrote a book on scalability talking about how he scaled a bank of MySQL servers.
What are the essential features of an enterprise system?
Support, availability of expertise, stability of the platform/language probably count.
But LAMP has other features like faster development, easier extensibility, lots of available libraries for reuse, several documented stories of scalability, maturing web frameworks.
Here are a couple of pointers to building Scalable systems (I am talking about Web Scale). I always wondered in the light of all this evidence, why the perception of LAMP as not being ready for Enterprise apps keep popping up.
As for Apache, every Netcraft study shows a very different adoption story. By the sheer number of servers, there may be more people with knowledge to configure, tune and extend the web server.
Scalable Web Architectures
Please Look at Market Share of all Servers Aug 1995 to Jan 2009
Linux is used a lot.
Apache and Tomcat are used a lot.
MySQL may be robust now. I'd use PostgreSQL instead. Banks will use Oracle, but there's good support for Java and Tomcat there.
PHP is used a lot, but many big companies would prefer Java.
You're best off arguing for a Linux, (possibly commercially supported version of) Tomcat, Java, Tomcat|Oracle|MSSQL solution, in my opinion.
You'll need a Linux sysadmin, especially as the number of servers ramps up, although I'm sure you can get a part time one in before that time arises. If the company already has Windows sysadmins then arguing for Linux is going to be tough.
I believe it's not that the technology is premature or something which keeps biggies like AT&T to go ahead with a full implementation at enterprise level. These companies have such a big budget for IT spends that the last thing they would have on mind is to spend more on the customization and enhancement required on the open source techs to suit their business needs.
So what they look for (which comes from my consulting experience) is buy and run product pack and don't have to spend more on the research and hack part. Companies which use open source build have developed their own support groups globally to cater to any support demands, which large enterprises are not much willing to do. They need thing done fast and for sure and they can pay.
There are two main issues for large enterprises using LAMP stacks:
TCO: taking into consideration that LAMP basically comes free, enterprises still achieve a lower total cost of operation with other commercial solutions
Supportability: enterprises have no problem paying the extra buck to get around-the-clock professional support from their commercial vendors
Redhat and IBM give full support for Linux, Sun bought MySQL, Yahoo uses Php, numerous companies use a LAMP stack, but many use parts.
I personally don't see Linux as being less well supported than the other OS mentioned; in fact hardware vendors typically DO support Linux over any other OS (except for Windows, which they do generally support quite well provided you use maintream distributions).
Provided you don't use a bizarre flavour (Tip: Just use RHEL or Centos which is its free equivalent), Linux is very well supported.
MySQL may have some shortcomings, but in my opinion it has many strengths; we use it at a large scale in ways not intended, but it still works quite well generally (most of the problems are due to our versions being out of date or badly configured).
What "P" stands for in LAMP is debatable. I feel that PHP is not enterprise-ready, because it has so many individual shortcomings (e.g. poor unicode handling, no namespaces, inconsistent APIs, inconsistent syntax, poor version backwards compatibility, duplicated/obsolete functionality) that they add up to making it difficult to implement a maintainable system.
But given an appropriately experienced team, even if you choose PHP it can be used to make an extremely high quality application.
If it's good enough for Google, trust me, it's good enough for you.
That's something I've been wondering for a while now.
Take Notepad++ for instace. Wikipedia tells me it was written in C++ and it's Windows-only.
Now take PHP. Wikipedia tells me this is also written in C++, but that runs on other OS too.
But I see more languages then just C++ for PHP... how is this done? Do they make some new code in C++, see it works and then figure out how to do it in Perl, or what happens?
It depends whether you are using platform specific libraries or not. Notepad++ is a desktop application and it needs a GUI toolkit. Although there are cross-platform C++ libraries like Qt and wxWidgets, Notepad++ is probably using a Microsoft's specific technology. Thus it can't be ported in other platforms.
PHP on the other side is a WEB scripting technology so there's no need of GUI library. Also there is much more stronger interest in running PHP in many platforms than there is for Notepad++. That is an incentive for the developers to make the C++ code cross platform.
Avoid platform specific libraries isn't the only thing needed for a C++ cross platform application. It usually means coding for the least common denominator and keeping different code branches for every platform supported. Although C++ is a cross platform language, each system has its own intricacies. In fact the code could be different in the same platform as well, if a different compiler was to be used. Try downloading the C++ source of an open source application, like PHP for example. You would notice that much of the code is the same for all platforms, but there would be different bits also. Sometimes preprocessor directives are used, elsewhere totally different source files are involved.
So creating a true cross-platform C++ application is a hard job and it is usually created when there is a strong incentive to do so and many people are involved. An one-man application like Notepad++ really can't be cross-platform.
Usually programs that work on a single platform make use of some facilities from the operating system (e.g. system calls to handle windows, buttons, services). Obviously the code is strictly system-dependent and cannot work in other environments. For cross-platform software you can follow several approaches, some of them are:
You program independently from the operating system and the stuff that lays under your feet (so you don't use operating system facilities, frameworks and so on).
You program using libraries and tools that are programmed that way, and thus work under several operating systems (for instance you use Qt or wxWindows, that are cross-platform, to manage windows and interfaces).
You build several versions of the code, to handle the particularities of the various architectures it will run on. This way you probably achieve more performance and "better" software because the code is optimized for its own architecture, but can be very difficult to maintain.
I think that most languages work in the first way, but sometimes they could use operating system calls to gain some performance (in this case you will have e.g. PHP for Windows, PHP for MacOS, ...).
Unlike some modern languages like Java and C#, C++ the language provides only basic functionality. It has no standard way of handling the user interface, threads, network interaction, cryptography, or even reading XML. Instead, support for this sort of functionality is left up to the operating system. Windows provides a broad API set called Win32 that applications written for Windows take advantage. There are similar APIs built on top of Linux and other operating systems. Whenever authors take advantage of the libraries of a specific operating system, they make it so their programs won't work on other operating systems.
It is possible to write a C++ program that will work across operating systems by abstracting away the items that interact with the operating system, but this is not simple and isn't often done.
C++ is an inherently cross-platform language; it's just that there are lots of libraries written in C++ that use features specific to a particular platform, so those programs are limited to that platform. Avoid those platform-specific features, and you've got yourself a C++ program that should compile on most platforms without too much hassle.
Notepad++ is Windows only because Notepad++ is a replacement for Windows Notepad, which is notoriously bad. No-one on any self-respecting OS would use an editor with 'Notepad' in the name ;>.
The core of Notepad++ is the Scintilla text editor component, which is cross-platform (Linux, OS X and Windows) and is used in both cross-platform and platform-specific applications. The core view logic is shared, with abstract classes for interfacing with windowing systems and graphics contexts. The applications take the cross-platform component and bind it to a OS specific window and graphics context. My favourite editor on Linux and Windows is SciTE, which is a very fast light wrapper around the editor with Lua scripting; there's also Komodo which wraps the Scintilla editor component in a Mozilla XUL container - another cross platform C++ stack.
There is some cost to making applications cross-platform; it's only worth that investment if there are prospective users on those platforms.
Apart from platform-specific libraries and APIs, C and C++ are also not truely cross-platfrom languages (for a good reason). They intentionally leave a lot of details unspecified, like type lengths, endianness, variable alignment in structs and whether newly allocated memory is initialzed.
This allows you to write (platform-specific) code that is as fast as your CPU can possibly go, but it also means that if you want your code to be portable, it takes some additional effort and testing - which many people targeting the Windows/x86 platform probably skip.
Others have commented on GUIs and use of other libraries that exist on only a subset of major platforms.
Another factor is the developers. Many developers (or software companies) only have expertise, access to, customer demand for, or need to use a single platform, and so they don't spend the extra effort to make their software cross-platform. For example, if a company has a Windows PC and developer tools on everybody's desk, no Linux or Mac machines in house to develop or test on, and no developers who are experts on those other platforms, and no large customers demanding a different platform, it's hard for them to justify not just plowing ahead with a Windows-only package. And if they change their mind later, when they have the expertise, equipment, or additional requirements, they may find it's too late to fix a large code base that has been allowed to become very platform-dependent.
Platform-independence takes real effort, every step of the way. If you start from day one with a plan, use cross-platform libraries (e.g., Qt, boost, OpenGL, etc., carefully avoiding MFC, DirectX, etc.), and build and test on all platforms regularly, it's probably only 10-20% more effort to make a good cross-platform app. But if you start with a one-platform app that's been in development for a long time, making it cross-platform can take as much effort as writing it from scratch, and that can be especially hard to justify if the new platform has a comparatively small market share in your industry or if your developers hate working on it.
Now take PHP. Wikipedia tells me this
is also written in C++, but that runs
on other OS too.
PHP is written in C only to make it available on more platforms. There are even rules to only make C style comments, not C++ //.
See PHP coding standards.