vicaya

24 ms·

211.
by vicaya 17y ago
Reuse is overrated and rewrite is underrated. It's true that typical first rewrite suffers from [second/sophomore system syndrome]( http://en.wikipedia.org/wiki/Second-system_effect ) The only way to overcome that is to do more rewrites :)
212.
by vicaya 17y ago
Well, this thing supports GUI/visualization as well, so many of these objects make sense.
213.
by vicaya 17y ago
Are you a manager?
214.
by vicaya 17y ago
Or if browser has builtin support (a la card space) for OpenID, which would solve the problem.
215.
by vicaya 17y ago
I'm fan of palm-fist salute (in martial art.) It predates Medieval but decidedly more hygienic :)
216.
by vicaya 17y ago
That there's some good in this world, Mr. Frodo... and it's worth fighting for.
217.
by vicaya 17y ago
MR tools, especially high level wrappers like cascading (which gives you capability to do easy joins on Hadoop) are very good for building indices for query. You can use these tools to process the (log) data once and load the results into a
218.
by vicaya 17y ago
You know what's more awesome than Mathematica? A free Mathematica with toppings (curated, structured data.) I hope they release a good API or just a formal query language, so I don't have second guess the NLP crap.
219.
by vicaya 17y ago
But no thank you.
220.
by vicaya 17y ago
Twitter is just one decent take on real time pub-sub systems, which has been around for a while, previous incarnations include mailing lists, usenet groups and various forums etc. Twitter simplified the pub-sub of short messages from/to mob
221.
by vicaya 17y ago
Raise hand, I'm a Hypertable developer.
222.
by vicaya 17y ago
General purpose malloc/free cannot "rule" in all situations due to its limited interface hence any particular understanding of resource management constraints. It's obvious that no malloc/free package can possibly outperform an arena scheme
223.
by vicaya 17y ago
Cuckoo hashing is a relatively new but underperforming hashing scheme due to the fact it's not cache friendly for modern processors: "despite its constant worst-case look-up cost, the bucketized cuckoo hash table was demonstrated to be the
224.
by vicaya 17y ago
Others should learn from this: free/unlimited inbound bandwidth encourages people to upload data into the cloud. Sounds like a nice platform for building crawlers.
225.
by vicaya 17y ago
It's impossible. Baiting titles are annoying.
226.
by vicaya 17y ago
In a lot of cases, just use forward declarations to avoid unnecessary includes in the header files goes a long way.
227.
by vicaya 17y ago
The best modern C++ numerical library I came across is Eigen: http://eigen.tuxfamily.org/index.php?title=Main_Page
228.
by vicaya 17y ago
If minor changes trigger long compilation, there is something wrong with your code organization. I found that if you count linking time the total build time difference between c and c++ for significant project is not that big at least not a
229.
by vicaya 17y ago
Yeah. "Overfit" is more of a ML concept than AI concept though. Many ML people don't believe in AI at all.
230.
by vicaya 17y ago
If you need range scan queries (selecting objects and their associations that match certain prefixes), the only DBs that can readily handle this are Bigtable (which is originally designed to store crawldbs) clones like HBase or Hypertable,
231.
by vicaya 17y ago
This is why I don't use Linux for desktop: atrocious font-rendering especially w.r.t to correct kerning: http://imgur.com/2bpz8.png (pay attention to the words: source, coined, success, recognition etc.) Only the Leopard gets it complete
232.
by vicaya 17y ago
Real time conversation/communication is overrated and has much lower S/N ratio than async email/comment etc.
233.
by vicaya 17y ago
Upvoted for Prof. Eugene Wang's perspectives. As a fan of Chinese calligraphy, I'm proficient in both Simplified and traditional Chinese characters.
234.
by vicaya 17y ago
Hypertable running in single node/local mode requires zero configuration and runs orders of magnitude faster than these alternatives when RAM to DB ratio is small.
235.
by vicaya 17y ago
One needs to deal with large amount of XML for non-pre-canned query. That's a fact.
236.
by vicaya 17y ago
Well, majority of the meat/bindings/scripts etc is written in XML. If you want to use something other than prewritten samples, you'll have to deal with XML. Period.
237.
by vicaya 17y ago
I'm surprised that neither HBase nor Hypertable are mentioned. Both HBase and Hypertable have much larger community and deployment than many of the mentioned alternatives.
238.
by vicaya 17y ago
uggh, yet another xml api. Either use JSON or something like Thrift or ProtoBuf.
239.
by vicaya 18y ago
> 8.9 File headers, revision history This is shocking on so many levels. People should let a VCS to keep the revisions.
240.
by vicaya 18y ago
The paper is written to sell Vertica, which is quickly becoming less appealing given the rise of open source parallel DBs like HBase and Hypertable. Hadoop is built for high throughput not low latency. If you read the paper, Hadoop runs cir

More ›