What do you think?
Rate this book
Understanding Compression: Data Compression for Modern Developers
If you want to attract and retain users in the booming mobile services market, you need a quick-loading app that won't churn through their data plans. The key is to compress multimedia and other data into smaller files, but finding the right method is tricky. This witty book helps you understand how data compression algorithms work--in theory and practice--so you can choose the best solution among all the available compression tools.With tables, diagrams, games, and as little math as possible, authors Colt McAnlis and Aleks Haecky neatly explain the fundamentals. Learn how compressed files are better, cheaper, and faster to distribute and consume, and how they'll give you a competitive edge.Learn why compression has become crucial as data production continues to skyrocketKnow your data, circumstances, and algorithm options when choosing compression toolsExplore variable-length codes, statistical compression, arithmetic numerical coding, dictionary encodings, and context modelingExamine tradeoffs between file size and quality when choosing image compressorsLearn ways to compress client- and server-generated data objectsMeet the inventors and visionaries who created data compression algorithms
242 pages, ebook
Published July 13, 2016
23 people are currently reading
66 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 4 of 4 reviews
January 27, 2017
I have a real problem with this book.
On the one hand, it does its primary job well. It sets out to be a quick and easy overview of the entire field of data compression so that the readers have some idea of the structure of a large field, and what may or may not be ideas relevant to their specific needs and interests.
If this were the entirety of the book, I'd give it four, maybe five stars.
On the other hand...
The book tries to adopt a jokey humorous manner throughout, and while I have no problem with that in principle, there is a real problem when the jokes are riddled with serious misinformation.
Let me give just two examples.
On the less serious side, there is a "joke" that mocks QuickTime as providing technology that was so stupid, har har, that all it could support was MJPEG files until YouTube was launched in 2001 and showed us all how to do things properly. Well, apart from the fact that YouTube was actually launched in 2005, and that QuickTime supported streaming (both live content and stored content) in 1999, this "joke" insults the entire QuickTime team (of which I was a member) by implying that we were all yokels who had no idea what we were doing. In fact QuickTime supported a variety of codecs from day one, and those codecs utilized, in various combinations, pretty much much all of the techniques described in the book.
On the more serious side, the authors make a horrifying mistake by claiming that entropy is a property of a set, rather than of a random variable or of a stochastic process. This then sets them up for making an on-going "joke" that those mathematicians with their entropy measure are so stupid that they can't even see how such obvious transforms as dictionary mechanisms, deltas, or RLE can be used for improved compression. Needless to say, this is complete nonsense --- even Shannon's original paper understood that the point was applied to stochastic PROCESSES, whereby the correlations between successive random variables has important consequences for the definition of entropy and thus for compression --- heck these various correlations and examples of how they play out are one of the standard features of the paper that's frequently quoted, and it appears early on, starting at page 5 with obvious examples on page 6 and page 7.
This matters. The greatest weakness in our current STEM education (IMHO) is the almost invisibility of measure-theoretic probability, which provides, in its language of measures, sigma-algebras, random variables, and stochastic processes, a remarkably powerful set of ideas for thinking about the world, every bit as valuable as sequence, derivative, and integral. But instead of being taught these ideas, most students (even physicists, CS students, and engineers) are taught a kind of "folk probability" that provides a few useful formulas but is essentially mired in confusion of the sort that this book so obviously displays, with no understanding of the difference between a random variable, a set of instantiations of a random variable, a random process, and its instantiations.
I hope the next version of this book tosses the "jokes" and instead includes a chapter (which doesn't have to be scary or complicated) simply laying out the basic foundational concepts of measure-theoretic probability at the same level of informality, and with the same goal of providing a useful overview, that's been applied to the actual compression algorithms.
On the one hand, it does its primary job well. It sets out to be a quick and easy overview of the entire field of data compression so that the readers have some idea of the structure of a large field, and what may or may not be ideas relevant to their specific needs and interests.
If this were the entirety of the book, I'd give it four, maybe five stars.
On the other hand...
The book tries to adopt a jokey humorous manner throughout, and while I have no problem with that in principle, there is a real problem when the jokes are riddled with serious misinformation.
Let me give just two examples.
On the less serious side, there is a "joke" that mocks QuickTime as providing technology that was so stupid, har har, that all it could support was MJPEG files until YouTube was launched in 2001 and showed us all how to do things properly. Well, apart from the fact that YouTube was actually launched in 2005, and that QuickTime supported streaming (both live content and stored content) in 1999, this "joke" insults the entire QuickTime team (of which I was a member) by implying that we were all yokels who had no idea what we were doing. In fact QuickTime supported a variety of codecs from day one, and those codecs utilized, in various combinations, pretty much much all of the techniques described in the book.
On the more serious side, the authors make a horrifying mistake by claiming that entropy is a property of a set, rather than of a random variable or of a stochastic process. This then sets them up for making an on-going "joke" that those mathematicians with their entropy measure are so stupid that they can't even see how such obvious transforms as dictionary mechanisms, deltas, or RLE can be used for improved compression. Needless to say, this is complete nonsense --- even Shannon's original paper understood that the point was applied to stochastic PROCESSES, whereby the correlations between successive random variables has important consequences for the definition of entropy and thus for compression --- heck these various correlations and examples of how they play out are one of the standard features of the paper that's frequently quoted, and it appears early on, starting at page 5 with obvious examples on page 6 and page 7.
This matters. The greatest weakness in our current STEM education (IMHO) is the almost invisibility of measure-theoretic probability, which provides, in its language of measures, sigma-algebras, random variables, and stochastic processes, a remarkably powerful set of ideas for thinking about the world, every bit as valuable as sequence, derivative, and integral. But instead of being taught these ideas, most students (even physicists, CS students, and engineers) are taught a kind of "folk probability" that provides a few useful formulas but is essentially mired in confusion of the sort that this book so obviously displays, with no understanding of the difference between a random variable, a set of instantiations of a random variable, a random process, and its instantiations.
I hope the next version of this book tosses the "jokes" and instead includes a chapter (which doesn't have to be scary or complicated) simply laying out the basic foundational concepts of measure-theoretic probability at the same level of informality, and with the same goal of providing a useful overview, that's been applied to the actual compression algorithms.
April 14, 2022
The first half is decent in explaining how some of the common compression algorithms work. The second half, though, feels like filler (there's even a one-paragraph chapter that only says the title content won't be covered). Sometimes the authors try to be too funny for their own good, to the point you are not sure whether one sentence is supposed to be satire or sincerity. Humour perhaps should be found elsewhere but books on a technical topic.
October 15, 2022
I've a joyful time reading this book. It show far more compression techniques than other text books, and save ton of time reading articles. I also enjoy when the book mention state of art like brotli, zstd, etc...
Displaying 1 - 4 of 4 reviews