Hacker News Reader

hglaseron Ask HN: What happens to older developers?

> FWIW, the few folks I know who tried to stay on purely technical paths ended up regretting it many years later.

Why?

hglaseron Bram Moolenaar responds to Neovim

Why aren't there other vim committers?

Bram's conservatism makes perfect sense for someone who has to maintain this huge, consequential, intimidating codebase. Why hasn't he gotten some help?

hglaseron Ask HN: Should I attend college?

Yes.

Yes yes yes yes yes.

"Front-end engineering" is a craft that happens to be relevant right now. It's an artifact of the particular client-server computing model we currently have. It is not likely to be relevant 5-10-20 years from now.

"Computer Science" is (among other things) a discipline that will give you the ability to learn the next craft, and the next one after that. It will keep you relevant your whole career.

Definitely go to college if you have the opportunity to do so.

hglaseron Ask HN: Where I can post my startup to get beta users?

This is super hard and there is no great answer. I've never met anyone who's had good luck with places like Betalist. (Though "Show HN" can work.)

Realistically you are going to be tracking down your first 100 users yourself manually. Figure out who among your friends is a good fit and bug them. Then ask all your other friends who they know who'd be a good fit. If they don't click, follow up and ask why. If they click but then don't come back, follow up and ask why. If they become an active user, you are about to become best friends, always talking about what they like and why.

You're going to feel like a mooch for a while -- like you're always asking your contacts for things and not giving back. This is normal.

Also, this thread is on the front page, so post a link. Quick -- an opportunity!

hglaseron Count distinct performance compared on top 4 SQL databases

We didn't try SSD drives. It's a good idea, worth a shot for a future post.

hglaseron Use SQL subqueries to count distinct 50x faster

OP here again. This was such a good idea, we decided to do it. :)

Here are the same queries tested on MySQL, MSSQL and Oracle as well: https://periscope.io/blog/count-distinct-in-mysql-postgres-s...

tl;dr: Oracle and MSSQL do very well, and even better on the more specific queries. Postgres's naive query is much worse than all the others.

hglaseron Use SQL subqueries to count distinct 50x faster

Good idea. Times are from my laptop, which is less beefy than the test rig we used:

  First query: 996s
  First query with your variation: 593s
  Second query: 67s
  Third query: 21s

hglaseron Use SQL subqueries to count distinct 50x faster

https://periscope.io/blog/atom.xml

hglaseron Use SQL subqueries to count distinct 50x faster

OP here. It's no EXPLAIN ANALYZE, but here are the EXPLAINs at least. The originals were averages of 10 runs. These EXPLAINs are just quick runs on my laptop, so no promises. :)

First query:

                                             QUERY PLAN
  -----------------------------------------------------------------------------------------------------
 Sort  (cost=3475858.60..3475859.56 rows=385 width=22)
   Sort Key: (count(DISTINCT time_on_site_logs.user_id))
   ->  GroupAggregate  (cost=3370756.94..3475842.06 rows=385 width=22)
         ->  Sort  (cost=3370756.94..3405784.03 rows=14010837 width=22)
               Sort Key: dashboards.name
               ->  Hash Join  (cost=44.90..558337.80 rows=14010837 width=22)
                     Hash Cond: (time_on_site_logs.dashboard_id = dashboards.id)
                     ->  Seq Scan on time_on_site_logs  (cost=0.00..365621.11 rows=14016911 width=8)
                     ->  Hash  (cost=29.40..29.40 rows=1240 width=22)
                           ->  Seq Scan on dashboards  (cost=0.00..29.40 rows=1240 width=22)

(10 rows)

Second Query:

                                             QUERY PLAN
  -----------------------------------------------------------------------------------------------------
 Sort  (cost=2709630.36..2709631.09 rows=291 width=26)
   Sort Key: (count(DISTINCT time_on_site_logs.user_id))
   ->  Merge Join  (cost=2604392.06..2709618.45 rows=291 width=26)
         Merge Cond: (time_on_site_logs.dashboard_id = dashboards.id)
         ->  GroupAggregate  (cost=2604392.06..2709521.80 rows=291 width=8)
               ->  Sort  (cost=2604392.06..2639434.34 rows=14016911 width=8)
                     Sort Key: time_on_site_logs.dashboard_id
                     ->  Seq Scan on time_on_site_logs  (cost=0.00..365621.11 rows=14016911 width=8)
         ->  Index Scan using dashboards_pkey on dashboards  (cost=0.00..87.00 rows=1240 width=22)

Third Query:

                                                   QUERY PLAN
  -----------------------------------------------------------------------------------------------------------------
 Sort  (cost=436381.86..436382.36 rows=200 width=26)
   Sort Key: log_counts.ct
   ->  Hash Join  (cost=436308.72..436374.22 rows=200 width=26)
         Hash Cond: (dashboards.id = log_counts.dashboard_id)
         ->  Seq Scan on dashboards  (cost=0.00..29.40 rows=1240 width=22)
         ->  Hash  (cost=436306.22..436306.22 rows=200 width=12)
               ->  Subquery Scan on log_counts  (cost=436302.22..436306.22 rows=200 width=12)
                     ->  HashAggregate  (cost=436302.22..436304.22 rows=200 width=4)
                           ->  HashAggregate  (cost=435705.66..435944.28 rows=23862 width=8)
                                 ->  Seq Scan on time_on_site_logs  (cost=0.00..365621.11 rows=14016911 width=8)

hglaseron Use SQL subqueries to count distinct 50x faster

OP here. Great points -- especially around MySQL and subqueries, which has been a frustration for us.

Comparing across different RDBMS's is a great idea. Maybe we can squeeze it into a future blog post. :)

hglaseron Tell HN: Planned Maintenance Jan 23 1-5 AM PST

Thank you for spending so much time on this. We appreciate it.

hglaseron Sequence of Events in the Cockpit on Asiana Flight 214

-third, there is an American instructor who wrote at Pprune (a pilot forum) about how for Korean pilots (trained in Korea), the worst emergency they can suffer in a simulator is a visual approach ... I can not find the article right now, I think it was erased shortly after being posted.

Here's a reprint. It's very interesting: http://www.freerepublic.com/focus/f-news/3041469/posts

hglaseron Paul Graham Shares Number of Y Combinator Companies Sold or Valued at Over $40M

Nope, YC's equity gets diluted like everyone else's. In fact, while many investors will buy more shares (at the new price) to keep their % ownership, YC does not, so their ownership % when the company exits is probably 3% or less, down from the original 6%.

hglaseron Show HN: The best way to run PHP apps

Really love these guys and their product. Congrats Kevin and Justin!

hglaseron $2,999 Mac Pro to go on sale in December

So what's the future of the MacBook Air at this point?

Does the Retina MacBook Pro just keep getting thinner until the Air is obsolete? Is there some fundamental technical reason they can't keep iterating the Air thinner and put a Retina display on it? Or a business reason why that's not a priority?

hglaseron Google announces Calico, a new company that will focus on health and well-being

Why is Google announcing this company? Will Google own it? Invest in it?

hglaseron Global Government Requests Report

Thanks for the CSV! Here are the top 10 countries by requests per capita[1]:

  Country               Requests    Population    Requests Per Capita
  Malta	                89          415,654       0.000214120398216
  United States (Min)	11000       313,900,000   0.000035043007327
  United Kingdom        1975        62,740,000    0.000031479120179
  Italy	                1705        60,720,000    0.000028079710145
  Australia             546         22,320,000    0.000024462365591
  New Zealand           106         4,405,000     0.000024063564132
  France                1547        65,430,000    0.000023643588568
  Germany               1886        81,800,000    0.000023056234719
  Singapore             107         5,184,000     0.000020640432099
  Portugal              177         10,560,000    0.000016761363636

Here's the full spreadsheet: https://docs.google.com/spreadsheet/ccc?key=0AiKnxhgG6s6VdE1...

[1] I used the minimum of the range for the US. Its ranking doesn't change if you use the max.

Population data is from Google searches, which lists the World Bank as a source.

hglaseron Global Government Requests Report

Thank you, edited.

hglaseron Global Government Requests Report

Re 3, couldn't it simply be that the US government has the most leverage over Facebook, since Facebook is headquartered in the US, and the majority of its staff and all its senior management live there?

hglaseron Global Government Requests Report

This is a great step. The analogous Google page seems to be here: http://www.google.com/transparencyreport/userdatarequests/co...

Can anyone confirm that these include National Security Letters?

CSV export would be nice. I immediately wanted to compute requests per capita by country.

hglaseron Life in the Boy's Dorm: My Career at Sun Microsystems

> The Valley is so much better now. There's no comparison.

How so? What's different? What's better?

hglaseron 37 YC Companies have valuations of or been sold for at least $40 Million

Why not use the valuation cap on the note (in cases where there is one) in place of the valuation? Is there a reason you wouldn't consider it equivalent to a valuation for the purposes of a rough calculation like this?

hglaseron Manchester City Council

What a predictable top comment for a Hacker News thread. :(

A government website showcases a design that is leaps and bounds ahead of any other government site we can think of, and HN's top comment casually dismisses it.

To rewrite something you said: "This is the perfect example of everything wrong with [HN comments]."

I, for one, am floored that someone in city government was able to rise above the bureaucracy and ship something truly useful and delightful for citizens. I would love nothing more than to meet the people involved and buy them a pint. Well done, guys.

hglaseron Cisimple partners with Kickfolio to make mobile app testing... simple

Continuous integration is huge, both for productivity and company culture. Nice to see mobile dev tools start to catch up with web tools. I'll be giving this a try for sure.

hglaseron Answering your questions about Heroku routing and web performance

Thanks for replying. As a paying Heroku customer (who's not affected by the routing issue), while seeing a blog post earlier would have been nice, it's great to hear that you spent so much time with affected customers.

hglaseron Answering your questions about Heroku routing and web performance

I would argue that the brand risk -- e.g. the risk of what ended up happening -- outweighs the legal risk.

That's not to say I wouldn't send it to a lawyer. But I'd do it with some hustle. Something like: "This is going out in 24 hours. Please comment ASAP!"

hglaseron Answering your questions about Heroku routing and web performance

This is a great response. I'm curious why something like this wasn't posted within 24 hours of RapGenius going public? I'd bet that a more thorough, technical reply would have mitigated a lot of the PR issues.

hglaseron Ask HN: Who is hiring? (April 2013)

San Francisco - Periscope - Outrageous Hacker!

Periscope makes data analysis tools that use probabilistic algorithms to achieve 100-1,000x real-world-measured speedups over existing tools.

We're well-funded by awesome investors, and the team is currently 2x ex-Google and 2x ex-Microsoft engineers. Come be our fifth full-stack hacker!

Email harry@periscope.io.

hglaseron Waiting for the Accelerator Bubble to Pop

If you had the ability to pick the top 46, why did you fund 84 in the first place?

(Not doubting, honestly curious.)

hglaseron Ask HN: Who is hiring? (March 2013)

Periscope

OUTRAGEOUS HACKER -- FULLTIME -- SAN FRANCISCO, CA

Periscope uses probabilistic algorithms to dramatically speed up data analysis. Our current algorithms achieve 13-225x real-world-measured speedups against up to 100M data rows. Come be our fourth Outrageous Hacker, working like crazy to make both those numbers go up while shipping the first version of the product!

We're well-funded by top VCs and our office in the heart of Soma is walking-distance from Caltrain and Bart. We're experience-agnostic as long as you're smart, get shit done, and excited to work at all levels of the stack.

Email harry@periscope.io

H1B Welcome