HNHacker News
TopNewBestAskShowJobs

vicaya

365 karma · joined August 24, 2007

hn@vicaya.com
submissionscomments
vicaya··on Ask HN: Does your lawyer attend your Board meetings?
This is a common practice for corporations, where you need to keep the meeting minutes as part of legal requirements of being a corp (c or s). You (as a Secretary of the corporation) can do it yourself or you can let them do it for you for free. Most people would choose latter.
vicaya··on Why Lisp is Awesome
Most of these "high-level" languages are more of "extensions" (CLOS certainly feels like an extension) or alternative lispish languages than bona fide DSLs with DS constraints (that can be used for DS optimizations). The S-expression syntax feels foreign/messy in many domains.
vicaya··on Why Lisp is Awesome
Risking being downmodded, I'd say Lisp is too "low level" for many common problems, which is great if your domain is parser/compiler construction/transformation/combination. I have no problem with lisp syntax (I can even tolerate C++ templates :), but it (and most general purpose languages) feels less natural for many domain specific problems than languages catering to the domains.

A DSL family based on a flexible core (with a lispish intermediate language even) might be a viable approach.

In many situations, simplicity and restrictions (constraints) in DSLs is better than flexibility and expressiveness in a general purpose language, because it can offer new possibilities for domain specific optimizations due to domain specific constraints.

Just like new notations in math led to breakthroughs in many math branches, new syntax can bring new life to software in many specific domains.

The key to such systems seems to be making DSLs easy to build and optimize using a suite of common building blocks (human (cs people) friendly intermediate languages (like lisp as opposed to VM byte codes) and libraries.)

vicaya··on Google should open source what actually matters: their search ranking algorithm
I'm an open source advocate (even for search algorithms in general), but ranking algo/model is inherently adversarial and gameable until true search AI (always correctly identify good and original content without false positives) is feasible.

I think it would be a good revenue strategy, if they charge $dollars to display top 10 positive features and top 10 negative features of any given query-document pair without any real weights/scores (maybe re-squashed with some logistic function). The $dollars amount is computed dynamically to offset the negative impact of the partial reveal of the ranking algo/model. This can be a tremendously useful tool for troubleshooting search results by site owners.

vicaya··on The patent system: End it, don't mend it
The problem is that despite the patent system, "trade secrets" remain and increasingly so in SaaS. Patents are mainly used as a weapon to impede the progress of competitors.

Abolish software patents, now!

vicaya··on The History of the Internet in a Nutshell
Hmm... where is the first web search engine? Google definitely didn't invent web search :)
vicaya··on Ask HN: need advice on how to write lots of data fast
Maybe my comment was not clear enough: buffering writes in memory on a single node alone has no durability to speak of. You need either a DFS or distributed DB that does replication of the buffers on different nodes to provide some reasonable durability.
vicaya··on Ask HN: need advice on how to write lots of data fast
It's generally true for traditional DBs, a few new generation DBs have options to buffer writes in memory and achieve durability via in memory replication in other nodes and only writes to disk sequentially in large chunks, which is better than simple logging as you'd have to sync to avoid data loss upon hardware failure.
vicaya··on The Most Intriguing Concept In Google's Go Language
Structural typing has been around for a while. Ocaml is a prime example.
vicaya··on I think I’m tired of Desktop Linux
The laptop is a desktop replacement 18.4" Acer with intel chipset (4500HD graphics etc.) bought from Amazon (free shipping and no tax! Looks like it's not available anymore, probably discontinued) 6 months ago. The netbook is a Dell mini 10v from Dell outlet on one of its 15% off offers.

The Acer has an HDMI output. I connected it to my 24" 1920x1200 Samsung and point and clicked the monitor applet on the gnome panel and viola, dual monitor support at full resolution without having to fiddle xorg.conf! Point and clicked to configure the wifi on both machines without any problems. Man, I was ecstatic, probably due to the low expectation from my Gentoo experience -- I was expecting to vi a few things to make them work :)

To be honest, I did have to add a fixmtrr.sh to gdm PostLogin and a line (MigrationHeuristic greedy) to xorg.conf to fix the intel video performance regression in Jaunty on both machines, when I felt the desktop effect was a bit sluggish. Now the desktop animations are smooth as butter even on the netbook.

I did choose to install 32-bit Ubuntu with server kernel (to get PAE in Jaunty) to take advantage of the full 4GB RAM, instead of 64-bit because flash and a few other things don't work as well as in 32-bit.

vicaya··on I think I’m tired of Desktop Linux
I've used a 2nd generation Macbook Pro that costs my employer close to $3000 (4GB RAM, 2.4GHz, 160GB HD) for 2 years, during which time the graphic card (Nvidia 9600M GT) died twice (screen suddenly went blank); battery died once (could not hold charge for an hour and would shut everything down without warning); Apple Mail corrupted its IMAP data countless times; much software annoyance: maximize/unmaximize/resize surprise etc.

Before the MBP, I've been using a $1000 amd64 notebook with 1.5GB RAM for 3 years without any issues, running Gentoo (! did need to google to fix multi-monitor support and ndiswrapper/wpa-supplicant for wifi) and it felt snappier than the mac with much faster processor running Leopard (up to 10.5.8).

Now I'm using a $600 laptop (with 4GB RAM and 320GB HD with higher display resolution than the MBP: 1680x945 vs 1440x900) and a 2.5lb $200 netbook running Ubuntu Jaunty. Everything worked out of the box including wifi, webcam and desktop effects. Feels much snappier than the Mac as well. All it took is a little googling _before_ buying the machines :)

It's much easier for Apple to QA limited configurations. Frankly, I'm quite pleased with the progress of Linux desktop.

Just another experience and perspective...

vicaya··on Google Go: A New Programming Language That’s Python Meets C++
The home page touts "safety" as a feature the language but all it takes to crash the program is concurrent access to the builtin map: "uncontrolled map access can crash the program", according the faq.

Interesting so far besides some syntax ugliness and lack of macros, but not yet good enough for many system programming projects, especially given the state of the garbage collector. GC is the hardest thing to implement large-scale, concurrent, memory intensive applications (e.g, database servers). Java spent 18 years to perfect its GC to be the current state of the art, which still sucks for these applications.

I hope they can come up with a concurrent garbage collector that can work on > 16 GBs of RAM with sub ms max latency :)

vicaya··on Ask HN: Review My Library, Tempest
Is there a way to change the colormap? Rainbow colormap is considered 'harmful': http://www.research.ibm.com/people/l/lloydt/color/color.HTM
vicaya··on Mike Stonebraker: The "NoSQL" Discussion has Nothing to Do With SQL
Actually many NoSQL DBs are not just about relaxing ACID requirements, but also about limiting functionality for easier scaling via DHT. They only support key/value set/get and not range scan, which is needed to implement SQL semantics.

I think the best interpretation for NoSQL so far is "Not only SQL", which has interface implication as well.

vicaya··on CrunchPad dead? Delayed by higher-than-expected costs?
Unlikely. Michael spoke before a small group of people less than a month ago, when I asked about CrunchPad. He said that it's coming along and that "Atom sucks." So, he's probably waiting for Chrome OS running on the new ARM processor (Cortex A9?) He mentioned that he has sunk $1M+ TC money in the project.

I'd love a bookpad that has two screens that fold together like a book. One screen is LCD that can serve as keyboard/(multi)touchpad/video screen and the other e-ink screen for reading books.

vicaya··on Lessons from Rackspace's downtime
When will the rest of the industry start to bundle a battery on every server, a la laptops and Google servers, so that PDU glitches like this would be no big deal?
vicaya··on Finally, semantic links for the girl i'm crushing on. what took so long?
Had a little cognitive dissonance when I noticed that the status is "accepted" for that attribute value.
vicaya··on The Future Is Big Data in the Cloud
All the traffic to/from your home is monitored. Encrypted data in the cloud is much safer and more available.
vicaya··on Stop calling me NoSQL
Native storage query language implies using native i.e., non-standardized query language and API. It's an categorical acronym like AJAX.
vicaya··on Stop calling me NoSQL
How about this 4-letter acronym: NSQL (Native Storage Query Language)?
vicaya··on Rewriting Playdar: C++ to Erlang, with code statistics
The stats are misleading, as it's not a comparable rewrite. The new code is refactored and much cleaner. For example, for "LAN Resolver 105 vs 427" a 75% saving. I looked at the actual code (thanks for posting it.) The C++ code contains a lot of cruft like explicit handling of ping/pong/pang and a http handler with some html snippets, which the new code doesn't have. I removed the cruft and result sloc is 105 vs 302 (still include all the boilerplate stuff: includes/namespace, the quite verbose asio server setup that can be refactored, and the line wasteful brace usage.)

The erlang code is also much denser. I gziped the code to compare a fairer symbol count: it's 1535b vs 2916b, a 47% saving, not bad but not that impressive either, considering the refactoring in the new code.

From user's point of view, the new distribution is 4x larger (10MB vs 2.5MB) and uses much more memory, which is a bad deal. It's OK for a developer to scratch his itch but dramatizing the benefit of such scratch doesn't help anybody.

vicaya··on Introducing BERT and BERT-RPC: GitHub's new serialization and RPC protocol
Any benchmarks compared with Thrift and Protobuf? BTW, Hadoop's new Avro project is yet another serialization project that's designed to be dynamic language friendly.

I personally find thrift IDL trivial to manage and is probably the only sane way to support a dozen languages (statically and dynamically typed) in a type safe and efficient way.

BERT-RPC is simple, but what if you find your server too slow and needs to be rewritten in another language? I think the Thrift approach makes that simpler.

vicaya··on "Design Patterns are the disease, Clojure is the cure..."
Well, although the syntax/idiom is different, the design pattern is the same the Observer pattern.

IMHO, design patterns transcend language idioms and enable people to understand/communicate higher level abstraction.

vicaya··on Every Developer Deserves More Memory
Yeah, I'd like to upgrade my brain with more and faster memory as well.
vicaya··on MooTools: The Open Source Magic Behind Cloudera Desktop UI
That's a fine reason, as long as you don't care about being friendly for integration/customization/embedding with external sources.
vicaya··on 80legs Web-Scale Apps Competition
Well, do you have something like this?:

http://setiathome.berkeley.edu/top_hosts.php

BTW, how do you ensure somebody is not submitting a malware (intentional or not) that can DDoS someone by crawling the same host at pretty much the same time?

vicaya··on NoSQL: Distributed and Scalable Non-Relational Database Systems
You're guessing the implementation here. It's only a bad idea if you're doing it wrong :) The source/conceptual tables can be normalized, while the join views can be materialized/implemented with Bigtable like sparse columns, which are dynamically updated if the joins happen a lot. Any subsequent join queries hits the views but the source tables. Manual denormalization should be allowed but not required.

GAE has nothing to do with this, it's a shared cluster, such that any benchmark is not reliable and comparable. It also has nothing to do with the SQL discussion, as it doesn't support joins either.

vicaya··on Riak - web-shaped data storage system
DHT can certainly support range queries by broadcast and merge. The keyword in the context though is efficiently. Scalaris not a typical Dynamo like DHT. It's based on Chord, so it's possible support range scan with O(logn) metadata RPCs (n is number of nodes.) per query, which I wouldn't call efficient as these RPCs are expensive in that they not very cachable (nature of the DHT, unless the range query is exactly the same.) Typical Bigtable implementation can easily cache portion of metadata with clients and resulting 0 metadata RPC in many cases.
vicaya··on NoSQL: Distributed and Scalable Non-Relational Database Systems
Denormalization can happen automatically based on the usage of queries with joins. Some commercial RDBMS like DB2 already do that. SQL is a declarative language. It doesn't force you to use a particular implementation.
vicaya··on MooTools: The Open Source Magic Behind Cloudera Desktop UI
Well, the facts mentioned in the post just confirms my judgment why MooTools/Prototype are bad: monkey-patching existing objects and polluting global namespace like no tomorrow, all in the name of object-orientation in a perfectly capable language that's prototype based and functional.
← PreviousPage 5 of 9Next →