It starts as the ultimate anti-structure. Nothing, not even MVC. But then you can refactor parts out of it and start moving towards a structure. Very freeing if it's a small one-off page/project and you know what structure works for you.
(Of course, if you need a database and authentication and Scaling and so forth, that's when you start bringing in the heavy artillery (libraries and packages), or switch to a framework.)
Malvina Reynolds - Little Boxes https://youtu.be/VUoXtddNPAM?t=14
It just keeps running and auto corrects any temporary load issues because of its stateless nature. I don't have to worry about memory leakage or processes that eventually become unresponsive and needs to be restarted. If the unlikely event that the PHP process crashes, it will just restart automatically and handle the next request as if nothing happened.
Its the ideal solution for side projects, and I am definitely going back now that I don't have time to maintain complicated infrastructure.
PHP was my generation's version of AWS stateless "lamda". Just place a file in the folder, and it will be executed on demand and clean up resources after the request has ended. A server could handle many different websites at the same time. That was basically how we could get cheap webhosting. They just put a bunch of Apache-instances behind a load balancer and used a shared network drive to read the files from
To be fair, it shows up as a nightmare from a certain scale. Toy projects might not reach this level..
Python has been around for like 30 years, why isn't that quite the case?
I guess there was mod_python but nobody cared enough: https://modpython.org/
I guess there's now mod_wsgi which is similar to PHP-FPM in some ways: https://modwsgi.readthedocs.io/en/develop/
Yet, even with the actual deployment not being much different (just used Apache as an example here), there's still dependency management, which Python is often described to be pretty messy at, virtual environments or not. With PHP, most of the stuff you need for common use cases often is pre-installed and is just a matter of enabling some plugin in the config.
That said, I deploy most of my stuff in containers, so PHP, Python, Node, .NET, Java and others all become more alike when viewed from the outside.
In the age of cloud, you just run your web application as its own server, and just have a proxy that fronts it.
How is a mainstream web server acting as a reverse proxy for an app server unsafe?
You get the benefits of one common approach to configuration, request logging (and possibly log shipping), rate limiting, compression, path rewriting and so on, while letting the app server handle the actual app and language runtime stuff (whether it's some Python app server, PHP, Java with Tomcat or whatever). You can even serve any static assets right there, alongside your API.
You can either include everything in a container, have separate containers for each process with overlay networking, or just have them run on the OS directly with only the web server binding to whatever ports you want to expose.
As for scaling, none of that prevents you from running anywhere between 1 or a million instances in parallel, with some load balancing (whether it's multiple servers or containers across a cluster). If it's really that bad, throw in HAProxy or something into the mix, alongside whatever cache solution tickles your fancy.
That's besides the point that I could provide a service to everyone in my country on a single rack (redundancy aside), not everyone even needs some mythical sky high scaling target.
> In the age of cloud, you just run your web application as its own server, a just have a proxy that fronts it.
Whether you use your own ingress (there's a lot of good ACME solutions out there) or something that your cloud provider gives you, using a server per app would in many cases be wasteful and there's a bunch of ways to increase the density, especially with the dynamic scheduling most container orchestrators could also give you.
mod_php was great for local iteration or when you wanted a very simple setup (and the same could be said about mod_python etc.), but most people have moved over to PHP-FPM nowadays: https://cwiki.apache.org/confluence/display/HTTPD/PHP-FPM There's probably more recent information somewhere, but this pretty much covers it.
Here's a very simple setup of two separate processes, with Supervisord (this is what runs in my self-contained dev container, but no reason why someone couldn't do something similar with systemd, or on a container cluster, with different users/permissions for each process):
[supervisord]
nodaemon=true
[program:php-fpm]
command=/usr/sbin/php-fpm -c /etc/php/fpm/php-fpm.conf --nodaemonize
[program:apache2]
command=/usr/sbin/apache2ctl -DFOREGROUND
LoadModule proxy_fcgi_module "/usr/lib/apache2/modules/mod_proxy_fcgi.so"
<FilesMatch \.(php|phar)$>
SetHandler "proxy:fcgi://127.0.0.1:9000"
</FilesMatch>
But yeah, I agree about the shortcomings of mod_php.
This achieves the same thing while providing more control and layering standard protocols.
php -S localhost:8000
> Python has been around for like 30 years, why isn't that quite the case?
> I guess there was mod_python but nobody cared enough: https://modpython.org/
Because it sucks! We tried it, it was bad, the Python community left it behind. Maybe if PHP left things that suck behind too then fewer people would say that PHP sucks.
> It is the purest form of web programming. You don't need a framework. Spin up a server and just start writing. Experiment. Do crazy things with it, like an art project. Every page is it's own world. Have fun with it.
Why would that be a bad thing? Even when it comes to something like REPLs, everyone always touts being able to iterate more quickly as a good thing, don't they?
Edit: aside from the implications of an in process language runtime, that was discussed in another comment.
“This is good because it enables you to iterate quickly” isn’t a good argument when you can iterate quickly regardless.
Necessary? Not quite.
But having those things be closer together leading to faster iteration? I'd argue that that's definitely true, same as how someone could skip out on writing a bunch of enterprise programming patterns, MVC or even tests, as well as use a language with a dynamic type system, or even write serverless functions. Not that any of that is very maintainable long term, but at the same time nobody would expect that from something that's very much like a REPL.
It's the same as how I like to have a clearly separated front end (typically a SPA) and a back end API (something RESTful) in most cases so that those technologies wouldn't be coupled and the API could be re-used as necessary, except that people who go for the various SSR and templating approaches typically code around me in circles, because they don't have to write as much API client code as I do.
Scripting? Ha! We used to do CGI in C when Perl wasn't fast enough. <old-man-yells-at-cloud />
How do you go into production? You put the file in a folder on the Web server. It works everywhere the same.
The code writing part itself is very straightforward and there's nothing counterintuitive as long as you don't try to do something that it's not designed for.
>there's nothing counterintuitive as long as you don't try to do something that it's not designed for.
Maybe that's the key. Instead of trying to write programs with tools that are not made for this but are somehow bullied into doing it, just use the right tools.
For generating HTML based on a request and data in the database, PHP it is. Don't try to make a language made for DOM manipulation process data and generate HTML.
For Web UI, DOM + JS it is. Learn how browser works and write the JS that will modify the DOM as needed which will result in interactive UI. Don't stack abstraction.
For example, both ASP and ASP.NET can be developed using Notepad.exe directly on an IIS web server, and you can similarly test changes by simply refreshing your browser.
Personally I'm allergic to this style of development, the "banging on it with a rock until it starts moving" approach tends to result in poor outcomes over the long term.
I much prefer that a compiler go over the entire codebase and check everything when I hit the compile button.
Sadly, this bad habit from PHP still lingers in ASP.NET Core to this day, and causes me endless grief.
Just over the last couple of months, I have encountered the following:
- DEV, UAT, and PRD are out-of-sync with each other and the source code. When I asked the developers why that is, they said they make changes to individual files instead of "building the whole thing", because "that's faster". Okay, they saved the two minutes the pipeline takes to run, great. Now I have to spend a week disentangling the mess.
- I refactor something, hit build, it comes back with all green ticks, I deploy... and the site is broken with HTTP/500 a day later because these scripting-style frameworks don't build the page code until runtime.
- A security vulnerability code scanner was enabled, and it broke the build. Why? Because it actually builds the code, all of it! Except that these frameworks have two sets of page template files that contribute to the "final product": 1) the list of page templates registered with the build system or VS project file, and 2) the list of files physically in the Git repo. Some build steps consider only the former, the web server considers the latter. You can remove a page from a build system, and it'll remain working. It's like removing C code from Linux and still have it turn up in the kernel! That's nuts, and a security nightmare. Cough-xv-cough.
PS: It's possible for ASP.NET Core to be configured to both compile every page on publish and support hot-reload and automatically include every file in the build if it's physically in the folder structure. This really ought to be the default, but isn't... to pander to ex-PHP devs or something.
I was sold on this too, that's why I left PHP for sexier stuff and never regained my productivity. As it turns out, most of the time I don't have a complete idea on what I'm doing but instead I have a direction and banging on it with a rock until it starts moving *in the right direction* is very productive because the design happens at the same time with the development.
My PHP code was complete "garbage" but it worked as expected and made me money and/or satisfaction of building things. Later I improved my coding skills and become much more aware of patterns and structures and found myself optimising the code instead of the product, destroying my productivity because as it turns out computers don't actually care or reward you for beautiful code. All that matter is, what that code is doing regarding the expectations of the humans who benefit from it.
Example 1:
I developed internal tools on a single server. The way I would do it was to have a personal folder on the server that I could use for development. Then I would commit the files to CVS which was the used for "deployment" to the main "production" folder. This worked well for me.
Example 2:
We also had a customer facing website where this practice was not allowed. We had proper build pipelines that created RPM packages which was installed by hand on the production machines.
My point is that there were many ways of avoiding editing the files directly in "production", even back when PHP ruled the world.
PS: I saw a developer take out an 80K user enterprise because they tried to change the look and feel of a login page live on the prod server. They didn’t know any better because they were used to editing pages in live systems. That bad habit resulted in about a million dollars of lost work.
Oops.
Well, not really because you have this massive php.ini file with tons of settings and whatnot, and scripts not working due to that was always a common point of failure back when I was using PHP. Also: not having some extension, which are almost always in C and require server installation (which is quite different from many other languages, where the user can usually specify most or all of their dependencies).
Maybe all of this changed? I don't know. But PHP in general has been the most fickle and unreliable environments I've dealt with. Give me any random servers, and I find it very hard to tell if my PHP script will run.
That's of course for singe server applications. However, because everything is straightforward and built for purpose and not trying to pretend being something else, more complex systems were still easy to setup and manage.
The joy of coding PHP usually died off with frameworks like Symfony where they built complex abstractions for the pleasure of the "Enterprise" projects with many moving parts.
I also have to say that php.ini is probably one reason PHP became so successful in the first place, by having things like safe_mode and such so it could be easily used on shared hosts with ease. All the $2/month shared hosts were offering PHP, not Perl or Python, because of this (and mod_php).
But other settings like that automatic $_GET/$_POST importing ... yeah, that was a lot more misguided. Even error_reporting really should never have been a system setting: always let the code handle these sort of things. And even safe_mode is mostly outdated and useless now, because no one is doing shared hosting any more.
Unfortunately all cloud providers I know bank on JS. I like JS but when it is in the browser, on the server it’s a disaster.
Today that sort of thing is very rare; you will run in a virtual machine or container, and have an execution environment all to yourself.
PHP-FPM which has options to specify system user per pool is basically how restrictions on file_get_contents are enforced.
However due to all the shit from developers PHP slowly became a half assed copy of Java. Instead of understanding what it was.
Node is the new php
Mixing server side and front end code is bad news. I think following many PHP-MySql tutorials will result in SQL injection vulnerabilities. Not good.
The same tutorial with Django wouldn’t have the same problem because Django auto-escapes strings you dump into HTML. These vulnerabilities only exist in this tutorial because PHP treats its output as HTML by default not text, so you need to put in extra effort to be secure.
People who are new to PHP and want to learn how to use MySQL will search for “PHP MySQL tutorial” and they will find insecure garbage.
The fact that you can find good tutorials does not negate the fact that people do find terrible ones.
And we aren’t talking about an article, we’re talking about a tutorial. There’s a very big difference between the two, why are you switching? Tutorials are obviously vastly more important to beginners.
Let me remind you of the context:
> > Mixing server side and front end code is bad news. I think following many PHP-MySql tutorials will result in SQL injection vulnerabilities. Not good.
> i think you'd have to go back at least a decade to find tutorials that bad
This is something that is harming people learning PHP today, not the distant past.
> The same tutorial with Django wouldn’t have the same problem because Django auto-escapes strings you dump into HTML. These vulnerabilities only exist in this tutorial because PHP treats its output as HTML by default not text, so you need to put in extra effort to be secure.
I don't understand why you keep ruminating about Django, which is obviously irrelevant as a comparison. If you want to use Django as a point of comparison you'd need to compare with Laravel.
You wrote something about how "people" do web development in Python, I don't see the relevance to your decrepit comparison.
Everyone that starts building computer network services is dangerous for about 5-10 years regardless of what tutorial material they initially come across. If not longer, since it takes a long time to internalise the protocols, client platforms, server platforms, relevant network layers, common threat vectors, and so on.