Bob's Nerd Shit

Mostly reading lists, rants about computer science topics, and personal projects.

[rkozik1989]

B-Trees and Databases

For years I’ve always known that B-Trees were used in databases, but until I snatched a copy of Principles of Database Systems (printed in 1982) I didn’t really understand how. The book breaks down how data is physically stored on a system and builds on that concept to illustrate how databases work. To keep things brief, records are basically their own data structure and indexes are systems of efficiently providing access to them, and what really struck me is in thinking about it that way that B-Trees really are not implicitly the best data structure for a Database. The tradeoffs really aren’t worth it unless you have large volumes of hierarchical data. So in saying that, I think we need to think more about our data and our application’s relationship with it when we’re designing software. Because it could very well be that an off-the-shelf database system wouldn’t be as performant as a hash-based system you rolled yourself.

[rkozik1989]

Rocks And Hard Places

When your passion is your occupation and you work as hard as I do from time-to-time crazy offers will slide your way, and you have to decide, do you want to keep what you have or do you want to roll the dice and try something new? Right now I’m in a very fortunate position. I’m working under somebody who’s had a successful career in the software engineering field for nearly 40 years, and the folks who’ve previously worked directly under him have gone on to be successful at companies like Google and Facebook. And to sweeten the deal further we work at a relaxed enough pace where I have the time for my independent academic pursuits.

But there’s more to this world and my career than the Facebooks and the Googles of the world, and as it would turn out, I am unlocking doors I didn’t think were un-lockable without one the aforementioned companies on my resume. Deciding whether or not to turn down said opportunities is terribly difficult, stressful, and not clear cut in the least. But ultimately, I think for the immediate time being, with the current economic situation, that the smart move for me is to stay put. There’s quite a lot I can learn at my current role that will help me better a candidate at those types of roles in the future.

[rkozik1989]

Output from my TypeScript call-graph generator I wrote in Python and DotViz.

[rkozik1989]

Why Scrap Old Software?

Developed in 1970 means 50 years of fixing bugs and handling edge cases. You try and re-write that thing and the end result will never be as feature complete or error free as what you've got now. If there’s a scalability issue, a more logical approach would be to figure out how to divide the work up among multiple machines. Age just shouldn’t be a factor in any decision to re-write an application.

[rkozik1989]

My RedHat Systems Administrator study stack

[rkozik1989]

Personal Projects I've Completed Since Mid-March

A TypeScript call graph generator, Git change log generator, JavaScript SimTower clone,  Pokemon Card search engine, a website with 80k pages, Roland T-727 clone, and the LAN party box. "With your condition, I don't think you'll be able to keep up." Lol, wut?

[rkozik1989]

Boutique Object-Oriented Programs vs. Framework-Based Programs

So I took my current job because I wanted to learn more about object-oriented software developed without frameworks and the flexibility that goes along with it, and my biggest takeaway so far is that frameworks and software packages are just self-contained versions of the entire software development ecosystem. That often solve some problems well but also provide light versions of software that's better suited to being a standalone service: queueing systems, caches, job servers, etc. Basically here's what that means to me, you shouldn't use a framework or a software package like Salesforce or SAP if you want to develop software that stands the test of time. If you want software that will scale to handle hundreds of millions or billions of records, reads, writes, etc. and last 10-20+ years you need to delegate tasks to well-suited services, libraries, and have the flexibility to roll your own as is the case with sharding. Rather, I think frameworks work well for small, disposable applications and  software packages are best when used to customize an experience.

[rkozik1989]

Building a Component Architecture in PHP

On one end of the spectrum you have procedural PHP applications. You know, the ones where all the logic you need for a page is contained on said page. And on the the other side of the spectrum you have whatever the hell it is that I’m building... a Zope-like component architecture in PHP. Why is this useful? In Web Development you typically have a single major version of software released at a time, and while it is most commonly the case, it is not always the case, as can be the situation with Internet of things applications. Where you might have major versions of software for each hardware set you maintain. There’s 3 obvious ways of writing such an application that immediately come to mind:

Design a single monolithic application where features are only enabled depending on what set of hardware is being used.

Write boutique standalone applications for each set of hardware.

Use a component or service-oriented architecture where the application is built from separate projects that are wired and glued together.

The problem with approach number 1 is that any hacker or modder could enable premium features on cheap hardware, and with approach 2 you’re spending an untold amount of dollars to design multiple bespoke applications. Which programmers are then going to have to learn. Though that’s not to say that the final approach is without its detractors. Component and service-oriented approaches do not come with a lot of features inside the box and development workflows are often difficult. But nevertheless I feel it’s important to explore what a component-based PHP solution would like and how you might tackle some of its pitfalls.

[rkozik1989]

I made a browser game for Chrome using plain old vanilla JavaScript(no frameworks or libraries), CSS, and HTML. The gameplay(if you want to call it that) is loosely based off of city-building mechanics but done in the style of Sim Tower. You can play it here: http://w.alled.city/ and view the source-code here: https://github.com/Rkozik/kowloon-walled-city

[rkozik1989]

Conventional wisdom says a site will be indexed by Google within 3-4 weeks, but what happens if your new site had 76,000 pages? Well, unless the rate increases I’m looking at 3.4 years.

[rkozik1989]

PHP Hatred Is Misguided

While it’s true that the language is terribly inconsistent and isn’t architected particularly well I don’t think those things matter that much at the end of the day. Performance? You can scale a LAMP application out to hundreds of millions of users and billions of monthly reads/writes, and frankly at that scale you’re more likely to have data model issues than language bottlenecks. At that size your biggest problem is coming up with a sharding and caching strategy and a means to implement it to distribute that load. Don’t believe? Well that’s what is powering the website, Flickr, Facebook, Wikipedia, Wordpress.com, and many others like Friendster that no longer exist. Rather, I think why some many folks hate on PHP is partly no different than why people hate on Nickleback or Guy Fieri. It’s a meme and you earn a lot of fake Internet points and laughs for doing so. Another problem is this 2010 presentation from Facebook and this rant from 2012. Facebook only hit language bottlenecks when they hit an unprecedented scale, and that rant doesn’t acknowledge the fact that an organization run like a plate du jour isn’t sustainable. You need to have tools that can do multiple jobs passably well to be able to fulfill business requirements and maintenance tasks in a timely fashion.

PHP, in my opinion, has lost favor with the folks who make decisions (management) because of its operating costs. The language consumes a lot of CPU time so you need quite a few servers. Finding PHP developers who’re also linux developers is quite hard so you need developers and sys admins. It’s also a very unforgiving language if you write bad code. Which means more time fixing bugs and less time working on new features.

[rkozik1989]

MySQL Sharding Readinglist

Tumblr Massively Sharded MySQL

Tumblr Architecture

Flickr Architecture

Facebook Architecture

NetLog Sharding

Sharding Architectures

LiveJournal Architecture

DRBD

Scale Out Mysql by Partitioning