Forgot your password?
typodupeerror

Comment: Re:WTB Cisco Switch (Score 1) 284

by Dewin (#42686539) Attached to: Cisco Exits the Consumer Market, Sells Linksys To Belkin

I remember having a Linksys WRT54G with a legitimate hardware issue years ago

Okay, I bought the story at Linksys but lost it at WRT54G. Though I suppose Cisco could have botched the newer hardware revisions ;)

Our WRT54G is about... 9 years old. Still runs as our primary router with nary a hiccup. I managed to get a hold of 3 WRT600Ns as well -- the one that I've actually put dd-wrt on and put into service (with the intent of expanding our wifi coverage to the basement) can't actually hold a decent connection speed (speed from wireless device to router is fine, speed from router to WAN is fine as evidenced by an Ethernet-connected PC, speed from wireless router to WAN is abysmal...)

Comment: Re:battery life (Score 1) 64

by Dewin (#41376357) Attached to: Motorola's First Intel-Based Handset Launches In UK

My phone (a Samsung Vibrant, or Galaxy S1 if you will) charges from empty to full on 2 hours.

Doubling the charge time for double the capacity doesn't seem like a problem to me, since it usually charges overnight -- and it still leaves the option of a half charge in the same amount of time for the same amount of battery life I have now.

Of course, I sometimes carry one of these around, but that's mainly because tethering is a huge battery drain. Oh, and that+the phone easily fit in one pocket, with the Nexus 7 that is using the phone-provided wifi in the other.

Comment: Re:Range data types (Score 1) 146

by Dewin (#41312835) Attached to: PostgreSQL 9.2 Out with Greatly Improved Scalability

Before 9.2, I did this (for timestamp ranges only) using Jeff Davis's Temporal Extensions for PostgreSQL, which I've submitted a few patches to.

Which really is the direct predecessor of 9.2's Range Type support (from the same developer, too.)

Heh, I should have guessed.

It was a few years ago, but we actually tossed around some ideas on a standard format for applying the range concepts to types besides timestamps. One of the issues then was that a small handful of built-in types have a notion of infinity/-infinity, but some (e.g. ints) do not -- yet ranges really need to support the notion, even a range of ints. It looks like the 9.2 implementation uses its own definition for -infinity/+infinity unrelated to the type in question, which thinking about it now might not have been the best decision (at least, without the ability to define that [,y] and [-infinity,y] are synonymous) since -infinity (as defined by range types) is less than all other values including -infinity (as defined by the contained type).

Comment: Re:Range data types (Score 1) 146

by Dewin (#41303373) Attached to: PostgreSQL 9.2 Out with Greatly Improved Scalability

Optimization of a constraint involving date ranges is a bit more difficult than you might think, and having it as one unified type makes queries a lot cleaner and indexes a lot more efficient (if done as GiST indexes anyways)

Old: WHERE (a.starttime BETWEEN b.starttime AND b.endtime OR b.starttime BETWEEN a.starttime AND a.endtime)
New: WHERE a.timerange @@ b.timerange

The speedup when you're doing things like trying to find overlaps between two lists of tens of thousands of ranges each is phenomenal.

Before 9.2, I did this (for timestamp ranges only) using Jeff Davis's Temporal Extensions for PostgreSQL, which I've submitted a few patches to.

Comment: Re:Simple solution (Score 1) 408

by Dewin (#40938541) Attached to: Secret Security Questions Are a Joke

The better designed systems use a one-way hash of the answer. The support guy types the answer in and it's hashed and compared. Other systems use a mix of reversible and hashed answers. It can be done securely.

Until your security answer is something that can be spelled and/or punctuated a multitude of different ways and you're answering it verbally (i.e. to authenticate yourself over the phone). "Grey" and "Gray" are going to hash differently, and to say nothing about the multitude of ways of spelling various names. ("Is it O'mally or O'malley or o'Malley or...")

You can't count on the support type spelling it the exact way you do.

And you definitely can't count on the back end to be storing it with any sort of security at all, so my original point still stands.

Comment: Re:Simple solution (Score 2) 408

by Dewin (#40934101) Attached to: Secret Security Questions Are a Joke

Yup. I had an embarassing phone conversation with my state's tax department because a year earlier I set the secret question to "What is the password?" and a year later I had naturally forgotten the answer.

This is a bad idea, since security questions are probably stored unencrypted or at least using a reversible cipher -- the people on the other end of support need to be able to compare your answer, and there needs to be some leeway especially with spoken answers and spelling variations.

Unless, of course, your answer is an entirely different password...

Those who can, do; those who can't, simulate.

Working...