revertts

Not correct: http://www.slideshare.net/AmazonWebServices/closing-keynote-...

What I gather from the slides:

1. server loads the data from S3 on start (sessions are sticky to a particular server)

2. user data gets mutated on that server for the lifetime of the session

3. the data gets written back to S3 on session end

How are failures handled in 2? If a box goes down, do you not only lose the open connections but also any data that has been changed in that session? If these are game sessions it seems like they'll be long-lived; do you periodically dump back to S3?

To me, a lot of this divide is between books I read cover to cover and books I consider references or pick up for very specific sections (like "Algorithms on Strings, Trees and Sequences").

CLRS, TAOCP, and Gang of Four are all books I haven't read in their entirety, but have been indispensable for certain topics. I still recommend them to people all the time, and feel perfectly justified doing so.

I found this book to be quite good: http://www.umiacs.umd.edu/~jimmylin/book.html

http://www.amazon.com/Data-Intensive-Processing-MapReduce-Sy...

Here's the pdf of it: http://www.umiacs.umd.edu/~jimmylin/MapReduce-book-final.pdf

This is nothing new. A number of large companies avoid GPLv3 because the potential of losing IP is too high; it's a difficult license to interpret, and the legal implications of some of its clauses haven't been fully defined yet (eg. what all is covered by "convey"?).

So, I wouldn't say that them avoiding it suggests anything sinister like some of the posters are mentioning... they're probably just playing it safe.

Seattle, WA. Full time

We're Amazon's platform team for media content. We own the systems for uploading, storing, and serving images/video/js/css/etc. across Amazon.com and partner sites.

Small blurb about the team: http://wap-jobs.amazon.com/jobs.html#media Full job description: https://us-amazon.icims.com/jobs/157798/job

Feel free to email me if you've got any questions!

Any chance you could link to it? I'd like to read the previous discussion, but I'm having trouble finding it on HN...

Do I get any control over node types, or is "number of nodes" the only knob I can turn?

There are also a lot of _really_ interesting but less public-facing teams (AWS and Kindle tend to get a lot of press). I had been pretty set on AWS when I accepted, but ended up in a different org and love the work I'm doing. You might talk to your recruiter or anyone you know at Amazon about your interests and see if they can suggest additional orgs/teams to consider.

If you did an internship, you're in a better position than anyone else to decide if the company's a good fit...

This might be too business sensitive to answer, but what did you replace BDB with as your disk backend? Build your own, LevelDB, or other...?

BDB certainly has its problems, but I still haven't found a better alternative.

I stopped reading at: There’s No 3G Web Browsing (except on Kindle Fire)

Kindle Fire doesn't even have 3G. They probably should've spent more time researching first.

Seattle, WA - http://wap-jobs.amazon.com/jobs.html (email is on the page linked)

Nearly every team in the Website Application Platform org is currently hiring. We build the backend components that power amazon.com and many of the subsidiaries/affiliates. We solve difficult problems involving scalable, performant distributed systems and have a direct effect on customer experience. If that sounds like you're cup of tea, it might be worth checking out!

Also, arcade cabinets, nerf gun fights, and free cereal. It's a good place to be. :)

I think stating it that way trivializes the amount of work that _has_ gone into lambdas. There's been a lot of other, more fundamental things discussed over that 2-3 year period about their implementation; way more than syntax.

This is pretty awesome. Food's the main reason I follow Groupon et al.

Yeah, that caught my eye too. I assume they just got Hadoop mixed up with Google's C++ MapReduce implementation.

For anyone wondering about the "back to 1" part: the Chinese very recently (this week?) took the spot with their 2.5 petaflop Tianhe-1A. It's made from Xeons, Fermis, and Feiteng-100s. Right now they're having trouble actually utilizing the machine; very little of their software takes advantage of more than a modest number of cores.

I'm really disappointed to see the language support for collections get pushed back to Java 8; it's one of those minor things that really starts to wear on me when I'm using the language.

I might just switch more code over to Scala...

Keyboard trays. You can't beat a good keyboard tray.

Thanks, that helped a lot. I had been trying to repeat with "."; hadn't noticed how many different kinds of repeats there are.

I prefer swapping Caps and Ctrl. I use Ctrl-[ instead of hitting Esc. It's an extra button press, but Ctrl is a more generally useful key (screen, C-c, etc.) in that location.

Last I checked there was no vi mode, which is why I stuck with zsh.

This is similar to what I use, except offlineimap instead of fetchmail and esmtp instead of ssmtp.

I like offlineimap because it does a bidirectionaly sync, meaning I can keep a local backup of mail, but still use the web client when I need to (without a ton of old mail sitting around).

I started using esmpt because it handles local mail delivery when paired with procmail, allowing me to get the infrequent messages to root without having to run a full MTA on my laptop. There's also a script to allow mail queuing, but I haven't ever bothered with it.

I used Gmail for quite awhile, then started looking for alternatives. Inbox.com is decent, but has no IMAP, which made it worthless for me. GMX Mail also looks like it could be good; seems very similar to Gmail. I hear very good things about Fastmail, but usually from paid users; if you're using a free account, they only give you 10mb for storage.

I eventually settled on paying ~$20 a year to Tuffmail. The features (Bayes classifier, sieve filtering, company who actually cares about my mail, etc.) made the marginal price worthwhile for me. If I were to go back to a free account, it'd probably be Gmail; I haven't seen anything better than it yet.

How To Solve It is a great recommendation.

I had a physics professor that did an amazing job of teaching problem solving; even though the class wasn't directly to my major, it was probably the single most useful class I ever took. Next semester I read How To Solve It and realized that the way he taught followed closely to what Polya lays out.

"Facebook is all php, while Twitter is built on Ruby on Rails."

Actually, Twitter's backend is Scala now.

I've been reading Cryptonomicon (Stephenson), Code Complete 2 (McConnel), and Algorithms: A Functional Programming Approach (Rabhi, Lapalme). Those first two are well-known and commonly recommended; the last one I just happened upon recently, and have really enjoyed it. It's the first algorithms/data structures text that I've seen targeting functional languages.

I've been meaning to read Godel, Escher, Bach (Hofstadter) for awhile now, but I'm not sure if/when I'll get around to it. Next on my list are some introductory number theory texts.

I'm going to have to disagree with '6. Keep Your Keybindings Standard.'

When I started with emacs, there were some bindings that would make my hands ache after a day of use, usually from the transition between difficult keybindings, not any one key combination in isolation. One of the first things I did was add in the ergoemacs bindings, and I never regretted it. It's not as if doing so made me suddenly forget what the old bindings were, although I am slower during the rare cases when I'm on the normal keybindings. But it's a small price to pay for not having my fingers hate me.

Editors are meant to be customized to your workflow and what works best for you.

Spideroak impressed me. It's a service similar to Dropbox, but with more emphasis on privacy. I found their devs really accessible and interesting to talk to.

This is largely why I switched my email to a paid provider that I trust plus local backups; it's simply too important.

Which is not to say that Google doesn't offer excellent services or that people shouldn't use them, but I think many people put far too much trust in a single provider without planning for things going wrong. Stories like this make me slightly apprehensive about all the buzz surrounding Google Wave.