Well, no. I'm indicating that antialiased fonts tend to be easier on the eyes, and the various UN*X web browsers tend to be particularily big offenders
Actually, the main problem tends to be with web pages designed for IE - it has a much larger default interpretation for the HTML relative font size scales (if you turn on large fonts on the Windows desktop, "larger" becomes "enormous"), and many of these sites use named Microsoft fonts for which a typical Unix setup (or at least, NS4.7 on RH6.x:-) substitutes a crappy small and non-scalable font.
Pragmatism and laziness on the part of page designers is forcing IE5.x and not the W3C as the referent standard for the web. NS has shot themselves in the foot with NS6, it sucks. Let's hope the Mozilla folks adapt to this pragmatism.
Is this all in aid of bringing us yet another implementation of that wonderful, anti-orthogonal 16/32-bit dinosaur instruction set? If they spent half as much effort on Merced, or better yet, revving a really good design like PA-RISC, we might get somewhere.
PA-8200 -> 4 flops per clock sustained
PA-8700 -> 8 flops per clock sustained
Intel P4 on dual RAMBUS -> 0.14 flops per clock sustained
I used to work as a sysadmin and now run an ASP service, here's some thoughts from the trenches:
1. Just because you have a support contract for closed source doesn't mean things will get fixed. It means you can call the vendor and ask for help, and they may or may not prioritize a fix. Mostly, these support contracts are only good for having them explain stuff YOU don't understand, or help you work a problem that's particular to YOUR configuration.
2. Just because you have a support contract doesn't make things get fixed faster. I once found a DNS resolver bug in the HP-UX kernel which I could have fixed in 20 minutes given the source (it was one of those where the problem is totally obvious from the behvaiour), but it took three weeks to escalate it through HP's official support system and get a patch. Eventually, I got the developer on the end of the line and explained it to him, and it was fixed in 3 minutes. This was as a large enerprise customer, and I wouldn't characterise it as a bad support experience, it's just the way the world works.
There is a bug in Word 97 (original and SR-1) whereby it doesn't use the standard filename handling library which the rest of MS-Office does and therefore can mess up UNC filenames which map to NFS or Novell IPX mounts. I was (I believe) one of the first customers to report this in the context of NFS, once again as a corporate client with 10k+ Windows dekstops, and although the level 1 people didn't find it in their database (it was listed as a Novell-specific issue) the support response from MS was excellent (escalated to Redmond in a couple of days) - they eventually gave me a workaround using a (then) undocumented registry setting. Had I had the source, I could have just browsed through and seen the code to use that setting, easy fix, no code to write.
Eventually, of course, we solved the problem permanently by switching from NFS on the Windows side to Samba on the server side.
3. Just because you have a support contract doesn't mean you have someone to sue. Look at the number of successful lawsuits against software vendors / support vendors for bugs. There are almost none. The only thing the support people are required to do is make a reasonable effort, and sometimes, that doesn't run to fixing the particular bug that upsets your installation and no-one else's (if it was a more widespread problem, they'd have caught it in QA).
4. The best support, for open or closed source products, comes from your peers - other sysadmins at other shops who are customers of the same vendor.
Even at the client level, where does the bulk of hours spent on Windows 98 support come from? Not from MS or Dell; it comes from neighbours, friends, nephews, etc. That's part of the secret to MS's success, they have over 10m unpaid part-time support people in the USA alone.
If you have any in house expertise at all, open source is far preferable, because you have the ultimate resort of debugging it / fixing it yourself. In all the time I have been using them in a business environment, I have never patched a single line of code in either Linux or FreeBSD's kernels, but I have in many applications, both proprietary (where I have obtained source as part of the license agreement) and pulbic domain open source.
In the proprietary case, I have rarely had time to work a bug myself, but I have had vendors provide me with 3 or 4 line patches over the phone which I could simply apply myself and then run make, much more expedient than a patch or point release.
In our ASP production environment, the only infrastructure pieces we DON'T have source code to are the back end DB platform (Solaris and Oracle). They are both products which are very mature and stable, and we're not pushing the envelope of either system, so I have little fear. We are OTOH pushing the envelope with Apache JServ, and I have taken the opportunity to rummage in the source on more than one occasion to understand what's under the hood, though I have yet to touch it myself.
In summary:
The theory that you'll end up patching source yourself is pretty thin, but you may well end up looking at it, and it can be damn useful.
Just as with software, free informal support like Usenet and mailing lists is often more useful than the kind you pay big $$$ for.
The quality of support at a pragmatic level depends more on the popularity of the product than it's development and licensing model. A fix is a fix whether it comes from Usenet or a vendor.
The sole difference between open and closed source from a support perspective is that you have the added benefit in the former case of having the source; the idea that closed source is somehow better supported because there is a formalised corporate structure around ownership of the IP is totally nuts.
We're on a UUnet T1 at work, and run our own mail server. I really don't want the hassle of having to bounce all my mail through UUnet's boxes, and neither do they - it just increases their network and server load for no reason. I pay them for a pipe and some minimal DNS replication. I *want* to run my own boxes.
Also, one thing the proponents of all these fancy ISP-based solutions seem to overlook is, does the ISP have the staff? With the competitive pricing in the market it is hard for ISP's to afford good staff. I won't mention names, but we had a customer whose office internet and email (and hence DNS) was provided by a major regional ISP, and when we put their new website live I went through three calls and about ten people to find a support person there who understood the concept of why I wanted an A record for the domain name itself.
The censorship seems to be restricted to US English usage... just tried NSI and it quite cheerfully offered to register (with advance apologies to our French speaking readers) encouler.com, and while poofter.com is taken, it suggested pooftercentral.com which is available.
Also, what about that famous site parodying Fast Company magazine and cataloguing the demise of silly.com startups? They must have registered somewhere.
The potential for computer performance that our technological society is capable of seems to have been sacrificed on the altar of Windows compatibility. All this clock speed, poured into a 32-bit RISC core emulating a totally broken CISC architecture.
All this clock rate for a measly 210 MFlops? Give us a break. HP's PA-8200 did over 700 MFlops sustained LINPACK in 1997, and at 200MHz.
So, if I install in on my wife's laptop and my desktop, so I subscribe twice? I don't subscribe twice to cable, or the newspaper.
But, since you are the sole person actually honouring your current MS user-serf licence, you have paid for two licences for Office for her right now - haven't you?:-)
Re:Check out the menu items at Venture Frogs...
on
Hacking The City
·
· Score: 1
We have an interesting demographic in our company:
The product the engieering team is working on is 100% Java server-side web stuff, using Apache and JServ, with dev environment scripting in Perl; developers do not have to support their own databases (IT group does that off a Sun box), and the parts of the production infrastructure that require Unix are also centrally supported on the office network, so it is easily possible to work on any platform and toolset, and we offer a choice of NT (currently 4/2000) or Linux (currently RH6.2/RH7). The only common ground tool requirement (apart from a Java compiler) that is mandated is CVS.
The production system runs on a multi-tiered Unix-only cluster (FreeBSD, Linux, Solaris).
About 1/4 of the engineering team use Linux. Most use it "pure" with no VMware fallback. I use it with VMware, which I use for PowerPoint (StarPresenter sucks) and IE (for comaptibility testing)
About 1/4 of them use Windows, with a Windows style environment, e.g. JBuilder and File Manager.
Most of the other half use the Windows OS, but with a purely "Unix" environment based on the GNU toolset; emacs, make, bash, textutils and binutils, jdb inside emacs, etc.
The resons cited for the last choice are varied, but basically boil down to availability of gadgetry - Quake runs better (before the flames start, these are basic office Dell boxes with 2D video); MS-Office is better integrated to Outlook/IE than StarOffice is to Netscape; there are a million and one little things you can download that just point and click on Windows, and most of them do work on NT.
At my old job I once tried to create the "GNu on NT" setup for myself, but I'm much more of a power user of Unix stuff so I ended up reinstalling the machine with Red Hat (note the specific name and not Linux generically; the toolset that comes with a good distro is 3/4 of what makes Linux powerful).
It's widely accepted (probably even here) that Windows is better for the computer illiterate because of the compatibility infrastructure around it; what I think is new here is the fact that it appeals to developers too, who aren't so dependent on that infrastructure.
No-one in America understands the current system - I have yet to find a US citizen who can tell me how the electoral college members are chosen, what precisely their constitutional mandate is, and the circumstances surrounding the times they ignore the will of the people (it has apparently happened on 9 occasions).
One of our guys at work used to work for Carl Rove and *he* doesn't know for sure either.
While I think the new legislation is great (guess who just switched jobs), it isn't quite enough to make things fair for all H1B
holders. Most (or at least a very large fraction of) H1B holders also apply for a greencard. If you switch jobs during the 4-5
years of the greencard process, your entire application is nullified.
No longer true - once the Labor Certification is complete, the rest of the green card process is transferable.
The market for high tech labor in the US is so tight that people have a freedom of choice - even the bottom end of the market represents a salary which offers a comfortable lifestyle, so people are making choices on other things. If companies treat staff badly they'll vote with their feet.
We are a startup, but we don't have kids fresh out of Stanford pulling 100 hours a week - we have mostly mature engineers with 10+ years of experience working hard (50 hours/wk or more, varying during a cycle) but balancing work with other things in their life. They produce more in that amount time than the college kids who live in the office, and at a higher quality.
At the last company I worked for (Austin TX, the locals will recgonize it) almost all the developers are 24 year old kids. You can SMELL the dev floor at the end of a release cycle, and it isn't pleasant. Their products are of woeful quality. Staff are heavily overworked, but paid looney salaries; kids 2-3 years out of college make up to $250k. But, after a year or two they wise up and quit anyway - there's more to life.
It is true that IT people in the USA generally work a lot harder than in other countries, but they are also well rewarded. There are jobs which are more laid back here in the US, or have more controlled hours - look at Siebel, who lock the doors at 6:30pm.
The real point of marvel about this quote, which most people miss, is not the fact that they could have undeestimated the potential market so much, but the fact that they invested in developing a product anyway.
I can't comment on Dell's service to individuals, I have always been a corporate customer. My home machine is a white box from a Linux boxbuilder, what can I say:-)
Dell quality - you are talking out of your hat. I have had exactly three hardware problems in several hundred machines at my last 3 companies; one fuzzy focus monitor, one power supply, one HD with stiction. All replaced immediately without quibble, based on a single phone call.
In terms of hardware support, yes they did have an annoying habit in the past of switching chipset suppliers, but that has settled a lot. Bottom line, you can shove in a Red Hat (or even FreeBSD) CD and it installs and supports all the hardware out of the box. Other parts of very acceptable quality - drives, memory, even CPU's:-)
Cases - Dell cases and motherboards are proprietary, but are designed for ease of service, not thumb-busting. You can open a Dell and get to all the parts in seconds. Card cage some out with a single pull in most of their boxes.
The mini-desktops with the laptop style CD-ROM drives make great tradeshow demo servers - we run them with 0.5Gb RAM, Linux and two network cards.
No comparison to a whitebox; it took me 10 mins just to shoehorn a second hard disk into my mini-tower AOpen case at home.
Support - this is where Dell really scores. I have had both new (with the full 3 year warranty) and used Dells (90 days parts only). I have only had first class service, and I have never had a Dell tech (Galway or Round Rock) question my judgement.
YMMV, but I really do think you're spouting hackneyed Slashdot wisdom rather than personal experience.
Well, I do the buying for our company and yes, I am one of you guys (techie, not PHB); I'm senior mgmt and run Linux on my desk:-) We are a startup and $$$ are important.
My policy is to buy Dell returns from DFO for desktops (they are new, just no warranty) and yes, they all come with Intel CPU's (Dell's choice not mine). We are now getting P3-733's with 133 frontside as "used" Dells. They are single CPU boxes. The reasons for my Dell-only are quality, homogeneity, ease of maintenance.
The cost difference between AMD and Intel is not enough to materially impact this policy, as I'd need to go with a different vendor; most business class machines (most of my desktops are for developers) only come with Intel, Compaq's AMD boxes are all low end.
All our servers (production and office) are low end SMP capable, and are either USparc-III (Sun Ex50's) or 2-way P3 boxes (Penguin, Dell, IBM).
Where I'd really kill for a good AMD is in the laptop arena - we use midrange laptops, mostly for business people (sales and marketing) who don't really need performance. In keeping with our frugal policy, we don't get monitors but we do make sure laptops have decent displays (1024x768 TFT) and this means we are forced to buy Celerons, which suck.
We have one real el-cheapo laptop, a bottom of the range $1000 Toshiba which we use as a test console for the server rooms. It was mine before we got venture capital. It has an AMD, K6-3 I think. The screen sucks (800x600 DS) but it kicks a $2000 Celeron laptop for performance, and the power consumption is very modest.
If I could get a decent midrange laptop with a good screen and an AMD cpu, I'd buy one in a heartbeat; on both performance and battery life, AMD kicks butt in the mobile market.
Because every userland app on Win32, including MS's, stomps the system DLL's during install, and for a host of other reasons, it is necessary for the ordinary Joe Schmoe *user* to be Administrator on their box just to use it, never mind fine graining admin privileges! Show me a Windows shop where users don't have root for their own boxes and I'll show you a captive terminal environment.
As to fine graining admin on Unix/Linux, it's done with groups. RTFM.
In practice, no-one ever does delegation anymore, even on VMS, Unicos MLS (a Unix which does support fine grained permissions) or NT for that matter; gone are the days of the operator chick who loaded 9-track tapes for you.
How about an Internet Startup? Most of them are fun places to work and for the obvious reasons many of them run on a Unix platform. We're a pure Java shop, and agnostic in Engineering; over half the developers run NT because the Quake/UT drivers are better (yes, really) but the production kit is FreeBSD, Linux and Solaris.
Slashdot Mirror
Well, no. I'm indicating that antialiased fonts tend to be easier on the eyes, and the various UN*X web browsers tend to be particularily big offenders
:-) substitutes a crappy small and non-scalable font.
Actually, the main problem tends to be with web pages designed for IE - it has a much larger default interpretation for the HTML relative font size scales (if you turn on large fonts on the Windows desktop, "larger" becomes "enormous"), and many of these sites use named Microsoft fonts for which a typical Unix setup (or at least, NS4.7 on RH6.x
Pragmatism and laziness on the part of page designers is forcing IE5.x and not the W3C as the referent standard for the web. NS has shot themselves in the foot with NS6, it sucks. Let's hope the Mozilla folks adapt to this pragmatism.
But, both my Amigae (1000, 3000/040) have fans :-)
However, the number of shootings in the UK in a year is still less than the number in LA in a month.
Used Optiplex GXi from Dell Factory Outlet - same price, Pentium, 32Mb RAM, CD-ROM, sound, integrated 100-BaseTX, modem and 2.1 Gb HD, 90 day warranty
Read the sci-fi novel "Distress" by Greg Egan, which has a wonderful vision of future networking, wireless and video tech as a subplot.
Is this all in aid of bringing us yet another implementation of that wonderful, anti-orthogonal 16/32-bit dinosaur instruction set? If they spent half as much effort on Merced, or better yet, revving a really good design like PA-RISC, we might get somewhere.
PA-8200 -> 4 flops per clock sustained
PA-8700 -> 8 flops per clock sustained
Intel P4 on dual RAMBUS -> 0.14 flops per clock sustained
I used to work as a sysadmin and now run an ASP service, here's some thoughts from the trenches:
1. Just because you have a support contract for closed source doesn't mean things will get fixed. It means you can call the vendor and ask for help, and they may or may not prioritize a fix. Mostly, these support contracts are only good for having them explain stuff YOU don't understand, or help you work a problem that's particular to YOUR configuration.
2. Just because you have a support contract doesn't make things get fixed faster. I once found a DNS resolver bug in the HP-UX kernel which I could have fixed in 20 minutes given the source (it was one of those where the problem is totally obvious from the behvaiour), but it took three weeks to escalate it through HP's official support system and get a patch. Eventually, I got the developer on the end of the line and explained it to him, and it was fixed in 3 minutes. This was as a large enerprise customer, and I wouldn't characterise it as a bad support experience, it's just the way the world works.
There is a bug in Word 97 (original and SR-1) whereby it doesn't use the standard filename handling library which the rest of MS-Office does and therefore can mess up UNC filenames which map to NFS or Novell IPX mounts. I was (I believe) one of the first customers to report this in the context of NFS, once again as a corporate client with 10k+ Windows dekstops, and although the level 1 people didn't find it in their database (it was listed as a Novell-specific issue) the support response from MS was excellent (escalated to Redmond in a couple of days) - they eventually gave me a workaround using a (then) undocumented registry setting. Had I had the source, I could have just browsed through and seen the code to use that setting, easy fix, no code to write.
Eventually, of course, we solved the problem permanently by switching from NFS on the Windows side to Samba on the server side.
3. Just because you have a support contract doesn't mean you have someone to sue. Look at the number of successful lawsuits against software vendors / support vendors for bugs. There are almost none. The only thing the support people are required to do is make a reasonable effort, and sometimes, that doesn't run to fixing the particular bug that upsets your installation and no-one else's (if it was a more widespread problem, they'd have caught it in QA).
4. The best support, for open or closed source products, comes from your peers - other sysadmins at other shops who are customers of the same vendor.
Even at the client level, where does the bulk of hours spent on Windows 98 support come from? Not from MS or Dell; it comes from neighbours, friends, nephews, etc. That's part of the secret to MS's success, they have over 10m unpaid part-time support people in the USA alone.
If you have any in house expertise at all, open source is far preferable, because you have the ultimate resort of debugging it / fixing it yourself. In all the time I have been using them in a business environment, I have never patched a single line of code in either Linux or FreeBSD's kernels, but I have in many applications, both proprietary (where I have obtained source as part of the license agreement) and pulbic domain open source.
In the proprietary case, I have rarely had time to work a bug myself, but I have had vendors provide me with 3 or 4 line patches over the phone which I could simply apply myself and then run make, much more expedient than a patch or point release.
In our ASP production environment, the only infrastructure pieces we DON'T have source code to are the back end DB platform (Solaris and Oracle). They are both products which are very mature and stable, and we're not pushing the envelope of either system, so I have little fear. We are OTOH pushing the envelope with Apache JServ, and I have taken the opportunity to rummage in the source on more than one occasion to understand what's under the hood, though I have yet to touch it myself.
In summary:
We're on a UUnet T1 at work, and run our own mail server. I really don't want the hassle of having to bounce all my mail through UUnet's boxes, and neither do they - it just increases their network and server load for no reason. I pay them for a pipe and some minimal DNS replication. I *want* to run my own boxes.
Also, one thing the proponents of all these fancy ISP-based solutions seem to overlook is, does the ISP have the staff? With the competitive pricing in the market it is hard for ISP's to afford good staff. I won't mention names, but we had a customer whose office internet and email (and hence DNS) was provided by a major regional ISP, and when we put their new website live I went through three calls and about ten people to find a support person there who understood the concept of why I wanted an A record for the domain name itself.
The censorship seems to be restricted to US English usage... just tried NSI and it quite cheerfully offered to register (with advance apologies to our French speaking readers) encouler.com, and while poofter.com is taken, it suggested pooftercentral.com which is available.
.com startups? They must have registered somewhere.
Also, what about that famous site parodying Fast Company magazine and cataloguing the demise of silly
The potential for computer performance that our technological society is capable of seems to have been sacrificed on the altar of Windows compatibility. All this clock speed, poured into a 32-bit RISC core emulating a totally broken CISC architecture.
All this clock rate for a measly 210 MFlops? Give us a break. HP's PA-8200 did over 700 MFlops sustained LINPACK in 1997, and at 200MHz.
So, if I install in on my wife's laptop and my desktop, so I subscribe twice? I don't subscribe twice to cable, or the newspaper.
:-)
But, since you are the sole person actually honouring your current MS user-serf licence, you have paid for two licences for Office for her right now - haven't you?
So, are VC's supposed to be an aphrodisiac? :-)
We have an interesting demographic in our company:
The product the engieering team is working on is 100% Java server-side web stuff, using Apache and JServ, with dev environment scripting in Perl; developers do not have to support their own databases (IT group does that off a Sun box), and the parts of the production infrastructure that require Unix are also centrally supported on the office network, so it is easily possible to work on any platform and toolset, and we offer a choice of NT (currently 4/2000) or Linux (currently RH6.2/RH7). The only common ground tool requirement (apart from a Java compiler) that is mandated is CVS.
The production system runs on a multi-tiered Unix-only cluster (FreeBSD, Linux, Solaris).
About 1/4 of the engineering team use Linux. Most use it "pure" with no VMware fallback. I use it with VMware, which I use for PowerPoint (StarPresenter sucks) and IE (for comaptibility testing)
About 1/4 of them use Windows, with a Windows style environment, e.g. JBuilder and File Manager.
Most of the other half use the Windows OS, but with a purely "Unix" environment based on the GNU toolset; emacs, make, bash, textutils and binutils, jdb inside emacs, etc.
The resons cited for the last choice are varied, but basically boil down to availability of gadgetry - Quake runs better (before the flames start, these are basic office Dell boxes with 2D video); MS-Office is better integrated to Outlook/IE than StarOffice is to Netscape; there are a million and one little things you can download that just point and click on Windows, and most of them do work on NT.
At my old job I once tried to create the "GNu on NT" setup for myself, but I'm much more of a power user of Unix stuff so I ended up reinstalling the machine with Red Hat (note the specific name and not Linux generically; the toolset that comes with a good distro is 3/4 of what makes Linux powerful).
It's widely accepted (probably even here) that Windows is better for the computer illiterate because of the compatibility infrastructure around it; what I think is new here is the fact that it appeals to developers too, who aren't so dependent on that infrastructure.
No-one in America understands the current system - I have yet to find a US citizen who can tell me how the electoral college members are chosen, what precisely their constitutional mandate is, and the circumstances surrounding the times they ignore the will of the people (it has apparently happened on 9 occasions).
One of our guys at work used to work for Carl Rove and *he* doesn't know for sure either.
While I think the new legislation is great (guess who just switched jobs), it isn't quite enough to make things fair for all H1B
holders. Most (or at least a very large fraction of) H1B holders also apply for a greencard. If you switch jobs during the 4-5
years of the greencard process, your entire application is nullified.
No longer true - once the Labor Certification is complete, the rest of the green card process is transferable.
The market for high tech labor in the US is so tight that people have a freedom of choice - even the bottom end of the market represents a salary which offers a comfortable lifestyle, so people are making choices on other things. If companies treat staff badly they'll vote with their feet.
We are a startup, but we don't have kids fresh out of Stanford pulling 100 hours a week - we have mostly mature engineers with 10+ years of experience working hard (50 hours/wk or more, varying during a cycle) but balancing work with other things in their life. They produce more in that amount time than the college kids who live in the office, and at a higher quality.
At the last company I worked for (Austin TX, the locals will recgonize it) almost all the developers are 24 year old kids. You can SMELL the dev floor at the end of a release cycle, and it isn't pleasant. Their products are of woeful quality. Staff are heavily overworked, but paid looney salaries; kids 2-3 years out of college make up to $250k. But, after a year or two they wise up and quit anyway - there's more to life.
It is true that IT people in the USA generally work a lot harder than in other countries, but they are also well rewarded. There are jobs which are more laid back here in the US, or have more controlled hours - look at Siebel, who lock the doors at 6:30pm.
Will they be hacking Red Hat, SuSe et al next to get the Linux sources? :-)
The real point of marvel about this quote, which most people miss, is not the fact that they could have undeestimated the potential market so much, but the fact that they invested in developing a product anyway.
Who the hell moderated this up? Someone with no clue, for sure.
The Cray T3x (the large Alpha system) is not SMP, it's distributed MPP, a true parallel computer. The SGI Origin and IBM SP/2 are NUMA.
Intel's SMP bus sucks, to be sure, but no-one does SMP beyond 32 CPU's - for good reasons few modern Slashdot readers understand.
And this used to be a techie website. Sigh.
You just don't get it, do you?
I can't comment on Dell's service to individuals, I have always been a corporate customer. My home machine is a white box from a Linux boxbuilder, what can I say :-)
:-)
Dell quality - you are talking out of your hat. I have had exactly three hardware problems in several hundred machines at my last 3 companies; one fuzzy focus monitor, one power supply, one HD with stiction. All replaced immediately without quibble, based on a single phone call.
In terms of hardware support, yes they did have an annoying habit in the past of switching chipset suppliers, but that has settled a lot. Bottom line, you can shove in a Red Hat (or even FreeBSD) CD and it installs and supports all the hardware out of the box. Other parts of very acceptable quality - drives, memory, even CPU's
Cases - Dell cases and motherboards are proprietary, but are designed for ease of service, not thumb-busting. You can open a Dell and get to all the parts in seconds. Card cage some out with a single pull in most of their boxes.
The mini-desktops with the laptop style CD-ROM drives make great tradeshow demo servers - we run them with 0.5Gb RAM, Linux and two network cards.
No comparison to a whitebox; it took me 10 mins just to shoehorn a second hard disk into my mini-tower AOpen case at home.
Support - this is where Dell really scores. I have had both new (with the full 3 year warranty) and used Dells (90 days parts only). I have only had first class service, and I have never had a Dell tech (Galway or Round Rock) question my judgement.
YMMV, but I really do think you're spouting hackneyed Slashdot wisdom rather than personal experience.
14 was the limit in the previous range of DEC AlphaServer machines (8400 series) due to design constraints (max 7 CPU boards, 2 per board).
They now do up to 32 CPU's in full SMP (there was an article on here last week about booting Linux on a 31 CPU Alpha system
Well, I do the buying for our company and yes, I am one of you guys (techie, not PHB); I'm senior mgmt and run Linux on my desk :-) We are a startup and $$$ are important.
My policy is to buy Dell returns from DFO for desktops (they are new, just no warranty) and yes, they all come with Intel CPU's (Dell's choice not mine). We are now getting P3-733's with 133 frontside as "used" Dells. They are single CPU boxes. The reasons for my Dell-only are quality, homogeneity, ease of maintenance.
The cost difference between AMD and Intel is not enough to materially impact this policy, as I'd need to go with a different vendor; most business class machines (most of my desktops are for developers) only come with Intel, Compaq's AMD boxes are all low end.
All our servers (production and office) are low end SMP capable, and are either USparc-III (Sun Ex50's) or 2-way P3 boxes (Penguin, Dell, IBM).
Where I'd really kill for a good AMD is in the laptop arena - we use midrange laptops, mostly for business people (sales and marketing) who don't really need performance. In keeping with our frugal policy, we don't get monitors but we do make sure laptops have decent displays (1024x768 TFT) and this means we are forced to buy Celerons, which suck.
We have one real el-cheapo laptop, a bottom of the range $1000 Toshiba which we use as a test console for the server rooms. It was mine before we got venture capital. It has an AMD, K6-3 I think. The screen sucks (800x600 DS) but it kicks a $2000 Celeron laptop for performance, and the power consumption is very modest.
If I could get a decent midrange laptop with a good screen and an AMD cpu, I'd buy one in a heartbeat; on both performance and battery life, AMD kicks butt in the mobile market.
Because every userland app on Win32, including MS's, stomps the system DLL's during install, and for a host of other reasons, it is necessary for the ordinary Joe Schmoe *user* to be Administrator on their box just to use it, never mind fine graining admin privileges! Show me a Windows shop where users don't have root for their own boxes and I'll show you a captive terminal environment.
As to fine graining admin on Unix/Linux, it's done with groups. RTFM.
In practice, no-one ever does delegation anymore, even on VMS, Unicos MLS (a Unix which does support fine grained permissions) or NT for that matter; gone are the days of the operator chick who loaded 9-track tapes for you.
How about an Internet Startup? Most of them are fun places to work and for the obvious reasons many of them run on a Unix platform. We're a pure Java shop, and agnostic in Engineering; over half the developers run NT because the Quake/UT drivers are better (yes, really) but the production kit is FreeBSD, Linux and Solaris.