It's called wear leveling, and yes, the other CF suppliers have it too.
It does reduce wearout, but you should really put any frequently updated and not too important files on a ramdisk.
With my CF based machines, I just did a standard (but minimal) slackware install and then used "find" to locate any files that got touched after leaving the system on for a day. A startup script copies those files into a ramdisk and symlinks them back into the directory tree -- so I got a standard linux install with now wearout worries.
Corel, via WordPerfect office, has been pretty instrumental in the emergence of low-cost PCs. The OEM price of this package is insanely low (around $10 +/-) which lets the PC manufacturers sell at a lower price point than they could if they equipped the system with Microsoft products.
Dell, HP/Compaq and Sony all ship Corel/WP Office with their low-end consumer systems due to the cost advantage.
I suspect that this might be a motivation for someone in the VC community to consider buying them. Low-cost PCs are a growth market.
Same here... Went from about 100 spams a day being 100% blocked by blacklists (primarily relays.osirusoft.com and the lists it contains) with zero spam making it through to 2-5 per day.
All email is: - Sent directly (no relays) - Usually from always-on internet accounts at cable/DSL companies. - Either ads for Viagra, email virii, or strangely email with no payload
All the email has forged return addresses and the content (if and ad) is using HTML obfuscation.
The problem with this new technique is because the spam is spread around so many hosts the usually spam reporting/blocking methods are less effective.
With a single host acting as a spam firehose, within an hour it is usually blocked and millions of messages are prevented from being recieved.
With 1000s of hosts, only a few hosts are being blocked, not stopping much spam. Also, algorithms such as Spamvop.net's are defeated because they depend on the volume of email from a single host to determine if the host should be blocked.
The only countermeasures I can see to stop this are either:
A "fast block" option -- a single unmoderated report of spam trigging a block for say one hour, and if more reports come escalate the time the source is blocked.
or
More direct countermeasures -- using some sort of automated hacking tool to recapture the systems have hacked and repair it/close it down. This is of marginal legality, I would imagine, though I think given the intent is benign and there's the internet equivalent of a clear and present danger it might be justified.
References and indirect lookups?
on
Gnumeric Turns 5
·
· Score: 1
Anyone know what the current state of different reference systems (R1C1 and A1) are, as well as their use in the INDIRECT() statement? Last time I checked this was still broken, though there was some discussion about it being fixed in the CVS version.
Internally we use some pretty complex spreadsheets and a few break gnumeric, which is one reason excel is still around. (They sometimes break excel, too.)
I've moved quite a few of the smaller but still important items to gnumeric, and it's not let me down yet. And the XML file format is great -- in lieu of a real scripting API I've used PyXML and python to read the Gnumeric files and process things that way, and it was really pretty easy to do.
It's a very small number. I research this stuff for a living, and 2% of people with PCs even open the case to upgrade graphics cards each year. Building your own system was less than 1% of PCs a few years ago, and that's when you actually saved money with DIY PCs.
With system prices now, unless you have components to reuse, it actually costs you money to build it yourself (i.e. buying a prebuilt Dell is cheaper than the prices of the components bought individually in low quantities.) Of course there are other (better) reasons to build your own system than just the money involved.
I so wish this were true. My company tried to trademark our name at the federal level (already TM'd in state) and we burned more than $10K in legal fees due to FTD (yeah, the flower guys) protesting the "Mercury" in our name. FTD's Icon/mascot is the greek god Mercury, but is never called/identified as such, and it's a freakin' corporate logo for a flower company. After the $10K and many years of effort, we just gave up. However, I make a point of never using FTD florist as a result (as did the entire staff of the law firm representing us, who were stunned by the boneheads at FTD.)
SIS' x86 core came from Rise Technology. The original Rise core was a Socket 7 (i.e. Original Pentium / Pentium w/ MMX) CPU with a shallow 6-stage pipeline. The core didn't perform very well except they had a pretty good MMX implementation (the benchmarks they used were usually MPEG playback, since Rise did well there.)
IIRC, Rise's CPU did get Microsoft WHQL certification, which means they did the x86 compatibility part good enough.
No telling what the details of current SIS' core are internally, but I'm guessing by the performance that it's pretty much the same thing. SIS claims three integer units and a 100 MHz FSB, which is what the Rise core had. SIS licensed the Rise core in '99 and again in 2002, so there were probably some improvements made to the original core (probably the FPU, which also was anemic.)
Just to clarify since a few of you were wondering and we're the source of the information:
- It's 1 billion x86 CPUs (8086 thru P4, all flavors). No 8080x, i960s, Xscale, etc is counted.
- Intel's figure is 1 billion. AMD is about 200 million units of x86 for the same time frame.
Also, Intel never comes out and says what their own data shows, primarily due to reasons related to stock,the SEC and the competion. There were some hints that Intel probably reached 1 billion before external researchers thought they did, but nothing official.
Slashdot Mirror
Your interpretation is correct.
Actually, Transmeta's share was unchanged. The increase in "other" was due to VIA (who is many times larger than Transmeta, FYI.)
It's called wear leveling, and yes, the other CF suppliers have it too.
It does reduce wearout, but you should really put any frequently updated and not too important files on a ramdisk.
With my CF based machines, I just did a standard (but minimal) slackware install and then used "find" to locate any files that got touched after leaving the system on for a day. A startup script copies those files into a ramdisk and symlinks them back into the directory tree -- so I got a standard linux install with now wearout worries.
Anyway, I agree -- CF is the way to go here.
Corel, via WordPerfect office, has been pretty instrumental in the emergence of low-cost PCs. The OEM price of this package is insanely low (around $10 +/-) which lets the PC manufacturers sell at a lower price point than they could if they equipped the system with Microsoft products.
Dell, HP/Compaq and Sony all ship Corel/WP Office with their low-end consumer systems due to the cost advantage.
I suspect that this might be a motivation for someone in the VC community to consider buying them. Low-cost PCs are a growth market.
Uh, it probably *IS* Bob Lazar. Under the "Prospecting Tour" link it includes this:
"Radiation and Uranium", a hands-on laboratory class personally instructed by the well-known "Area 51" scientist, Bob Lazar.
A friend of mine says people on the coast in florida refer to this as "winning the lottery" or that they went fishing and caught a "square grouper."
(Grouper being a fish for the sea-life challenged among us.)
Same here... Went from about 100 spams a day being 100% blocked by blacklists (primarily relays.osirusoft.com and the lists it contains) with zero spam making it through to 2-5 per day.
All email is:
- Sent directly (no relays)
- Usually from always-on internet accounts at cable/DSL companies.
- Either ads for Viagra, email virii, or strangely email with no payload
All the email has forged return addresses and the content (if and ad) is using HTML obfuscation.
The problem with this new technique is because the spam is spread around so many hosts the usually spam reporting/blocking methods are less effective.
With a single host acting as a spam firehose, within an hour it is usually blocked and millions of messages are prevented from being recieved.
With 1000s of hosts, only a few hosts are being blocked, not stopping much spam. Also, algorithms such as Spamvop.net's are defeated because they depend on the volume of email from a single host to determine if the host should be blocked.
The only countermeasures I can see to stop this are either:
A "fast block" option -- a single unmoderated report of spam trigging a block for say one hour, and if more reports come escalate the time the source is blocked.
or
More direct countermeasures -- using some sort of automated hacking tool to recapture the systems have hacked and repair it/close it down. This is of marginal legality, I would imagine, though I think given the intent is benign and there's the internet equivalent of a clear and present danger it might be justified.
Anyone know what the current state of different reference systems (R1C1 and A1) are, as well as their use in the INDIRECT() statement? Last time I checked this was still broken, though there was some discussion about it being fixed in the CVS version.
Internally we use some pretty complex spreadsheets and a few break gnumeric, which is one reason excel is still around. (They sometimes break excel, too.)
I've moved quite a few of the smaller but still important items to gnumeric, and it's not let me down yet. And the XML file format is great -- in lieu of a real scripting API I've used PyXML and python to read the Gnumeric files and process things that way, and it was really pretty easy to do.
It's a very small number. I research this stuff for a living, and 2% of people with PCs even open the case to upgrade graphics cards each year. Building your own system was less than 1% of PCs a few years ago, and that's when you actually saved money with DIY PCs.
With system prices now, unless you have components to reuse, it actually costs you money to build it yourself (i.e. buying a prebuilt Dell is cheaper than the prices of the components bought individually in low quantities.) Of course there are other (better) reasons to build your own system than just the money involved.
I so wish this were true. My company tried to trademark our name at the federal level (already TM'd in state) and we burned more than $10K in legal fees due to FTD (yeah, the flower guys) protesting the "Mercury" in our name. FTD's Icon/mascot is the greek god Mercury, but is never called/identified as such, and it's a freakin' corporate logo for a flower company. After the $10K and many years of effort, we just gave up. However, I make a point of never using FTD florist as a result (as did the entire staff of the law firm representing us, who were stunned by the boneheads at FTD.)
SIS' x86 core came from Rise Technology. The original Rise core was a Socket 7 (i.e. Original Pentium / Pentium w/ MMX) CPU with a shallow 6-stage pipeline. The core didn't perform very well except they had a pretty good MMX implementation (the benchmarks they used were usually MPEG playback, since Rise did well there.)
IIRC, Rise's CPU did get Microsoft WHQL certification, which means they did the x86 compatibility part good enough.
No telling what the details of current SIS' core are internally, but I'm guessing by the performance that it's pretty much the same thing. SIS claims three integer units and a 100 MHz FSB, which is what the Rise core had. SIS licensed the Rise core in '99 and again in 2002, so there were probably some improvements made to the original core (probably the FPU, which also was anemic.)
Just to clarify since a few of you were wondering and we're the source of the information: - It's 1 billion x86 CPUs (8086 thru P4, all flavors). No 8080x, i960s, Xscale, etc is counted. - Intel's figure is 1 billion. AMD is about 200 million units of x86 for the same time frame. Also, Intel never comes out and says what their own data shows, primarily due to reasons related to stock,the SEC and the competion. There were some hints that Intel probably reached 1 billion before external researchers thought they did, but nothing official.