Note that if you want to use it the usenet way, you can. Just change the options.
Of course, it amazes me that a bunch of slashdotters even care about this stuff, because it only applies if you're doing HTML mail.
If you're using plain text email, then it all just works quite nicely, with a sane LIFO approach and no indents, no colors, no garbage. Just the information.
The problem is that its not how that typically works. You typically will have also gotten all the 15 replies and forwards in between, so you're already current when #16 comes in, and you dont need to re-read all the old stuff.
You talk about less work, but the least work is just to read whats on top, not have to scroll at all, and be done with it.
The only way I could see the other approach paying off is if you get alot of email in Digest form, rather than one at a time, or if you want to skip ahead and not read all the interim messages, just the final ones.
While Outlook's text editor isnt anything to write home about, you've got a couple things wrong here.
Outlook has two default text styles: "compose" and "reply." Assuming nobody bothers changing them, after the second reply everyone will be typing in the same font and color. Only if you're using HTML mail and default options.
Plus, as an added bonus, Outlook's quote is just an indent and a set of email headers. There's no nice ">" at the start of each quoted line or nice blue line like there is in Thunderbird. Only if thats how you want it. It's almost infinitely configurable.
Your choices are:
- do not include original text - attach original message - include original text with no formatting changes - include and indent original text - prefix each line with ? (pick what you want to prefix it with)
You can also choose whether to reply above or below.
Thats pretty damn flexible.
It defaults to a very reasonable default for most people.
For most folks, top-replying is the correct choice because it serves most of the people most of the time.
And, because as already mentioned, Outlook's email editor sucks, Outlook really doesn't handle inserting new lines of text into quoted sections that well. Assuming nobody's done anything fancy with formatting it will simply unindent the line of text. However, you'll still be typing in the blue "reply" format unless you've changed that style, so the only queue that it's a reply is that it's not indented. Unless you're the first reply after an email is sent, then by default you'll be typing blue and their text will remain black. But after one round, this is lost. Note that this kind of stuff only happens if thats how you have Outlook configured and leave everything default. And you have to be (crazily) using HTML email for this. It's all quite nice and sane when you've set outlook to do everything as plain text.
In short, it's because Outlook's email editor basically sucks. It wants to be an embedded Word instead of an email editor. Outlook's email editor IS Word, at leats for HTML and RTF messages. Literally it uses the MS Word engine for it. Quite alot of brouhaha a while back about that because it really limited the kind of HTML you can use.
For those who've never used Outlook, I've essentially formatted my post in a general "Outlook reply" format. Keep in mind that the quoted section would just be indented, without the little quote lines that Slashdot has added. Unless you've configured it to put quote lines like Slashdot has.
Keep in mind as well that for most people, top-reply works most of the time, thats why most email editors default that way. Otherwise for every single email that you read that has a reply you have to scroll offscreen to read it, instead of what you want to read being right there.
I will say that Thunderbird's default reply-formatting IS better than Outlook's. But its pretty minor, and you can make Outlook look very similar if you want to fiddle with the customizations.
It has not been updated in years and the single database for all of your information makes it little more than a toy. I'm not sure what you mean by has not been updated in years. In 2003 they introduced the Cached Exchange mode, which makes transitioning between online and offline transparent and invisible to the end user. Thunderbird is just terrible with online/offline support.
Lots of other nice minor improvements since the XP days. The only real complaint most folks I encounter have with Outlook is that you cant have your Calendar and your Mail as separate windows (so you can alt+tab between them).
As for the latter part of your statement, I think you're confusing PST's, which hardly anyone uses, with an Exchange back end, which is much more common.
Years ago, they had a device advantage for sync but Open and portable are all the rage today. Not quite. All people care about are 'does it sync' and how much does it cost them to suppor the sync. For anyone who has had to support executives with palmpilots and handhelds, the Windows based handhelds with activesync against exchange is light years better than the Palm software. The latter still requires you to run as local admin to work, for frick's sake.
you're assuming that all costs would be in addition to their current costs - highly unlikely. Actually its even worse than that, as they likely get revenue from T-Mobile now, so it would go from a (small) income stream to a large cost.
Third, you are assuming there are no offsetting revenues - if there aren't there is no point in offering the service in the first place. Any offsetting revenues they're already getting with T-Mobile, so no net change there.
Finally you are assuming no economies of scale which Starbucks clearly would have. Actually its the other way around. It's alot easier for a single mom & pop store to put up a crappy router & wap and just say, "If it works, it works." A company the size of starbucks cant easily do that. You are correct in that if they did build up a support environment then they could have scale effects but that would be expensive. Benefiting from economies of scale requires big up front investment.
There is NO WAY a wifi rollout would cost Starbucks $282 million (3% of 2007 gross revenue - your numbers not mine) That would be $33,000 per store. You're right, my 3-5% of gross was offsetting hyperbole to your 1% of profit not being material.
When the wap is starting to fail, or the barista rebooted the network equipment in the wrong order, or when there are too many people on doing p2p and it doesnt work worth a damn.
This happens all the time in the 'free' wifi places I've been to. Whereas with T-Mobile at Starbucks, it always works. It's fast, low latency, and I have someone to call if there is a problem.
I'm at a loss trying to decide who I have less sympathy for--T-Mobile for thinking they can charge $10 for a Wi-Fi connection, or Starbucks for thinking that providing the $10 connection is going to bring in the 'Net-savvy customer. You're not understanding the situation.
Hardly anyone at Starbucks pays per-day fees, with the rare exception of someone travelling.
The vast majority of users are on a per-month unlimited plan.
They're professionals and business people, who use starbucks and t-mobile all over the country to get their business done.
They're not trying to bring in 'the net-savvy customer'. They're trying (and succeeding) to bring in the business customer, cause he's got a business who will pay per month fees forever.
And you're just covering the cost of the monthly fees.
Plus hardware, plus setup, then maintenance, then a support system so people have someone to call when it doesnt work, etc etc.
It adds up fast.
The biggest reason why a company like Starbucks wouldnt do it is because its not their core business. It takes some infrastructure, specialized knowledge, and a whole staff to manage a network like this, keep it up, and provide end-user support.
So they'd have to build a whole internal department in their company to manage this, and it would end up being more like 3-5% of gross revenue.
For a big company, its not always acceptable just to setup a router & wap, wide open, and say 'have at it'.
You'll then have people get pissed off and leave when it doesnt work, plus you'll have customers bugging your staff about it, and they'll be trying to troubleshoot it.
It only takes a handful of questions a month to your staff about this before you've easily doubled your ISP fees in real costs.
You do realize that your case isnt typical, or even terribly repeatable?
Unless you've got a fantastic location in a high traffic area where people have money, coffee shops are fairly low margin.
Reading this post and the post below, you've got to be running at least $1000 a month in monitoring and consulting bills (equiv I know since you're doing it yourself).
But most shops dont have people like you in them, and hiring people (like me) is going to cost money. This level of cost will just wipe out most small coffee shops.
Most small shop owners struggle to remember to reboot the cable modem BEFORE the wifi unit, and dont touch it unless their customers complain.
And hell, most cities dont even have 16/2 available, or if they do, for anything less than $500 per month.
So yes, you've done some neat stuff, but you're a one-off.
Your experience and approach is not generally repeatable for other small coffee shop owners.
Sure it may not be the top 1% connoisseur level, but its a step above all the crappy mom and pop shops around.
The vast majority of independent coffee places make bland, tasteless coffee, and dont have a clue of the difference between good and bad.
At Starbucks, if you go it may not be the best of the best, but its usually going to be a cut above most things around it, and have a huge selection, insanely friendly and helpful staff, clean stores, and quality wifi.
If you go to the big cities, sure there are some top notch independents or small chains out there, but they're rare. Starbucks is in the top 10%, but there are better.
Mind you, many cities, like my home city, unfortunately, doesnt have anything better locally other than going to a high-end roaster yourself and making your own coffee. But who wants that.
it's ridiculous to think that people are going to pay another $30 a month for wifi. Alot of people do. T-Mobile made alot of money in that deal.
I pay for it (rather my business does) because its so damn convenient.
I also have a Sprint EVDO RevA card, but I tend to use the t-mobile at starbucks alot more because:
1. There's always one just right down the street.
2. The quality of the service is excellent, and its NEVER down.
3. It's much faster and lower latency than my sprint card.
All that being said, you're right in that its not a reasonable deal for casual users. But a business offering wifi at its stores is more complicated and expensive than you think, and there are tradeoffs at every point.
In my experience, a large percentage (ie, the majority) of 'free' wifi spots are down or so slow to be useless a significant amount of time. This is a hassle, and costs me.
Whereas if I go to a starbucks, its always up (4+ years with 100% service for me), and the speed and latency are excellent (1.5mbps synchronous, good latency).
This quality guarantee is worth the $30/month for many people.
You're right in that it DOES suck for the casual user though. Not everyone makes their living off their computer and so can justify spending the money.
However, for people like myself and others, the T-Mobile/Starbucks thing has been a godsend the past 5 years.
Quality coffee, with high consistency, and a high-quality wifi connection anywhere in the country that you go. Hard to beat it, from a business user standpoint.
In addition, the quality of service of the wifi at Starbucks is excellent.
I've been to so many 'free' wifi spots where the wifi was down alot, or just so badly managed and/or congested and/or oversubscribed that it was useless.
I'd rather pay a few bucks a month and have something that 'just works' than have to roll the dice each time I sit down at a coffee shop.
much cheaper than paying $40 a month per location for dsl/cable, assuming each store could even realistically get broadband service. I'm fairly sure this is not correct.
I've looked at my traces and such while in there, and it doesnt appear to be a cellular network, nor does it have the ping times and latency associated with it.
In addition, T-Mobile has been present in my city's starbucks for years before any sort of cellular data service was available here (that could reach 1.5mbps at least).
Next time I'm over there I'll take another look, but I dont think this is correct.
It's possible they do this for some sites where they cant get dsl, but not generally.
every place that has 'free' wifi, is a place where they put in high speed internet for their 'inventory' system, and the 'free wifi' piggybacks on that internet connection. in some cases, they use satelite for the inventory system ugh. I'm also fairly sure this isnt true. I've been to a large chunk of the major cities in the US, and have visited multiple starbucks in each one, and I've never seen a 'free' wifi at a Starbucks.
Even mirroring is basically useless for reliability if the drives are from the same vendor unless you swap out the mirrored drive daily Can we please leave the ridiculous hyperbole at home?
Although what you say sounds reasonable in theory, it doesnt work out that way in the real world.
Even people using multiple deathstar drives didnt see consistent failure times.
Very very few drive failures are because of defects. The deathstar notwithstanding. Even that though isnt relevant because no one in their right mind would use a drive like that for something important. I mean... it actually said on the box that it wasnt intended for 24x7 use.
Everyone that has released data and in my personal experience, hard drive failures are quite random and arbitrary. Even when a huge number of them is from the same manufacturer and model.
So please be careful what kind of mal-information you bring here, because someone might believe you and try to carry this bad information back to the workplace.
The risk level on RAID is quite simple. It's a ratio between expected drive failure rate and rebuild time.
There's a reason corporate systems still tend to build arrays out of 36, 75, or 150GB drives. It's because rebuild rates are reasonable (and also more spindles equals more speed). So if you ever see someone building a raid array out of 500GB consumer level drives, you know they are either:
1. Working with data that can take a much higher risk of multi-drive failure, since rebuild times are huge.
2. Working with multiple systems networked, and using replication between them to handle multi-drive failure risk.
3. Working with data availability needs that can stand going to tape if the whole array downs.
4. Idiots.
If your rebuild rate is more than an hour or so on very important data, then you use 10 or 6 or one of the higher raid levels.
So you plan what is the cost of losing a second drive while the first is rebuilding, and you design your storage around that.
Like everything else in IT, you pay for how much risk reduction you want.
In my experience, 2 drives failing in the same array close enough in time to lose the array is extraordinarily rare. However, sa user-error while trying to rebuild the array (ie, do it wrong and down the whole array, or pull the wrong drive) is fairly common.
If that happens, you go to your backups. If you cant afford the downtime by that, then you use a RAID level that you can afford and provides you the level of protection you need.
But RAID is not useless, and telling people that is just feeding bad information to people who dont know any better.
Then why is there an entire new dynamic language support infrastructure (ie, JSR292, invokedynamic) being built into Java 7?
My understanding is that JRuby and similar on the JVM use wrappers and conversions (basically a custom translation interface), whereas they are working on a more generically supported approach.
Some quick googling on jsr292 invokedynamic and java7 should give some more info.
I saw some of your stuff after I wrote that response, but appreciate you posting the links anyway.
I'd like to see a more thorough analysis of some of this stuff by someone more qualified by me, using actual debuggers and signal analysis on the bus.
However, in the meantime, your tests are fairly straightforward (also glad Ostermann added some useful points about IHV added effects and whatnot), and match my personal experience.
Because, while, in my experience, an unpatched XP machine (haven't tried this with Vista) is owned in an average of 43 seconds. 2001 called and they want their meme back.
This hasnt been an issue since SP2 came out on WinXP, and you know this.
He's a academic imaging expert with no concealed interests and someone who cared enough about Windows to investigate Vista. He never investigated Vista. He investigated some white papers, and some people who work at Nvidia and ATI.
He never (according to his website) investigated Vista. It's all second or third hand, or worse.
All of this because Gutmann dared publish a careful and objective review of Vista's DRM. Except thats not what Gutmann did.
Gutmann read some very old white papers. He then built a mental model of what Vista might look like if people implemented it according to his personal interpretation of those white papers.
He then assumed that mental model to be true, and slammed this theoretical model that might be Vista on his interpretation of how MS might have implemented it.
But read his website closely. He clearly, at least according to anything I can find on the website, never once tried to actually see if any of his theories were correct.
Which is really strange, because thats the fundamental scientific method. Postulate a theory, then test your theory. Gutmann only ever did the first part.
Everything Gutmann said has proved correct. Interestingly enough, no. There hasnt really been any refutation or support for any of Gutmann's claims, one way or another.
If you have a decent machine and good drivers, Vista clearly doesnt behave at the level of flakiness that Gutmann claimed it would, which seems to automatically invalidate some of his claims.
But the verdict is out, as far as I've seen. No one has ever attached hardware to the bus and tried to see if there are flip bits, or the kind of intrusive checking that Gutmann claimed would be in there.
It might be, it might not. No one that I have read has EVER tested it, including (as best I can tell from his publishings) Gutmann.
Plus, both require a huge, ugly back end to be installed. No they dont. You can serve silverlight from a vanilla apache box, with no mono installed on the back end.
The silverlight plugin is also quite small (a couple of megs).
I agree, though for a little bit different reason.
Flex/Flash IS a much strong platform now, plus all of it except the Flash player is open-source (blazeds, amf, swf format, flex sdk, etc).
However, give a few years, and Silverlight/DLR/XAML will be quite the hot thing, technically. You'll always be strapped to MS using it, but aside from that, the underlying technology design is superior, IMO. Silverlight and the DLR are just way too young at this point.
But look at some of the comparisons:
Silverlight: XAML, Flex: MXML
MXML has quite a legacy background in movie/animation flash. It's evolved from there to a more typical event driven application. XAML was developed from the ground up for this and you can see it in how easy some things are to do in it. But XAML is also way young, and missing some things that just arent implemented in this version.
Silverlight: C#, any DLR language, Flex: ActionScript 3
Good lord, not even a comparison here. Just about any modern language is better than javascript/actionscript. Mind you, AS3 is a big leap over AS2 and previous. But still. It's javascript at the core, which is never good.
All that being said, we're opting for Flex on the front end and java/spring/hibernate on the back end. Mostly though because Flex is ready now, and platform independence and not being tied to MS is important to us.
They did a quick release of Silverlight 1, which was pure XAML markup and javascript. Very light, very simple, and you could only do media-style apps on it really.
Silverlight 2 is in beta now, and is quite impressive technically, but very very young yet. XAML is quite nice, and being able to use C# or any DLR language in your silverlight code (as opposed to actionscript in flex/flash) is very very nice.
However, its very young, and wont have some traction for a while.
Flex/Flash isnt quite as sexy (technically) but has the advantage of alot more maturity, and less vendor dependency.
Slashdot Mirror
Note that if you want to use it the usenet way, you can. Just change the options.
Of course, it amazes me that a bunch of slashdotters even care about this stuff, because it only applies if you're doing HTML mail.
If you're using plain text email, then it all just works quite nicely, with a sane LIFO approach and no indents, no colors, no garbage. Just the information.
The problem is that its not how that typically works. You typically will have also gotten all the 15 replies and forwards in between, so you're already current when #16 comes in, and you dont need to re-read all the old stuff.
You talk about less work, but the least work is just to read whats on top, not have to scroll at all, and be done with it.
The only way I could see the other approach paying off is if you get alot of email in Digest form, rather than one at a time, or if you want to skip ahead and not read all the interim messages, just the final ones.
Your choices are:
- do not include original text
- attach original message
- include original text with no formatting changes
- include and indent original text
- prefix each line with ? (pick what you want to prefix it with)
You can also choose whether to reply above or below.
Thats pretty damn flexible.
It defaults to a very reasonable default for most people.
For most folks, top-replying is the correct choice because it serves most of the people most of the time. And, because as already mentioned, Outlook's email editor sucks, Outlook really doesn't handle inserting new lines of text into quoted sections that well. Assuming nobody's done anything fancy with formatting it will simply unindent the line of text. However, you'll still be typing in the blue "reply" format unless you've changed that style, so the only queue that it's a reply is that it's not indented. Unless you're the first reply after an email is sent, then by default you'll be typing blue and their text will remain black. But after one round, this is lost. Note that this kind of stuff only happens if thats how you have Outlook configured and leave everything default. And you have to be (crazily) using HTML email for this. It's all quite nice and sane when you've set outlook to do everything as plain text. In short, it's because Outlook's email editor basically sucks. It wants to be an embedded Word instead of an email editor. Outlook's email editor IS Word, at leats for HTML and RTF messages. Literally it uses the MS Word engine for it. Quite alot of brouhaha a while back about that because it really limited the kind of HTML you can use. For those who've never used Outlook, I've essentially formatted my post in a general "Outlook reply" format. Keep in mind that the quoted section would just be indented, without the little quote lines that Slashdot has added. Unless you've configured it to put quote lines like Slashdot has.
Keep in mind as well that for most people, top-reply works most of the time, thats why most email editors default that way. Otherwise for every single email that you read that has a reply you have to scroll offscreen to read it, instead of what you want to read being right there.
I will say that Thunderbird's default reply-formatting IS better than Outlook's. But its pretty minor, and you can make Outlook look very similar if you want to fiddle with the customizations.
Lots of other nice minor improvements since the XP days. The only real complaint most folks I encounter have with Outlook is that you cant have your Calendar and your Mail as separate windows (so you can alt+tab between them).
As for the latter part of your statement, I think you're confusing PST's, which hardly anyone uses, with an Exchange back end, which is much more common. Years ago, they had a device advantage for sync but Open and portable are all the rage today. Not quite. All people care about are 'does it sync' and how much does it cost them to suppor the sync. For anyone who has had to support executives with palmpilots and handhelds, the Windows based handhelds with activesync against exchange is light years better than the Palm software. The latter still requires you to run as local admin to work, for frick's sake.
What do you do when the wifi isnt working?
When the wap is starting to fail, or the barista rebooted the network equipment in the wrong order, or when there are too many people on doing p2p and it doesnt work worth a damn.
This happens all the time in the 'free' wifi places I've been to. Whereas with T-Mobile at Starbucks, it always works. It's fast, low latency, and I have someone to call if there is a problem.
Hardly anyone at Starbucks pays per-day fees, with the rare exception of someone travelling.
The vast majority of users are on a per-month unlimited plan.
They're professionals and business people, who use starbucks and t-mobile all over the country to get their business done.
They're not trying to bring in 'the net-savvy customer'. They're trying (and succeeding) to bring in the business customer, cause he's got a business who will pay per month fees forever.
Thank you!
It's amazing how no one else around here gets this.
1% of annual profit is NOT a rounding error.
Thats material and noticeable.
And you're just covering the cost of the monthly fees.
Plus hardware, plus setup, then maintenance, then a support system so people have someone to call when it doesnt work, etc etc.
It adds up fast.
The biggest reason why a company like Starbucks wouldnt do it is because its not their core business. It takes some infrastructure, specialized knowledge, and a whole staff to manage a network like this, keep it up, and provide end-user support.
So they'd have to build a whole internal department in their company to manage this, and it would end up being more like 3-5% of gross revenue.
For a big company, its not always acceptable just to setup a router & wap, wide open, and say 'have at it'.
You'll then have people get pissed off and leave when it doesnt work, plus you'll have customers bugging your staff about it, and they'll be trying to troubleshoot it.
It only takes a handful of questions a month to your staff about this before you've easily doubled your ISP fees in real costs.
You do realize that your case isnt typical, or even terribly repeatable?
Unless you've got a fantastic location in a high traffic area where people have money, coffee shops are fairly low margin.
Reading this post and the post below, you've got to be running at least $1000 a month in monitoring and consulting bills (equiv I know since you're doing it yourself).
But most shops dont have people like you in them, and hiring people (like me) is going to cost money. This level of cost will just wipe out most small coffee shops.
Most small shop owners struggle to remember to reboot the cable modem BEFORE the wifi unit, and dont touch it unless their customers complain.
And hell, most cities dont even have 16/2 available, or if they do, for anything less than $500 per month.
So yes, you've done some neat stuff, but you're a one-off.
Your experience and approach is not generally repeatable for other small coffee shop owners.
Compared to what.
Sure it may not be the top 1% connoisseur level, but its a step above all the crappy mom and pop shops around.
The vast majority of independent coffee places make bland, tasteless coffee, and dont have a clue of the difference between good and bad.
At Starbucks, if you go it may not be the best of the best, but its usually going to be a cut above most things around it, and have a huge selection, insanely friendly and helpful staff, clean stores, and quality wifi.
If you go to the big cities, sure there are some top notch independents or small chains out there, but they're rare. Starbucks is in the top 10%, but there are better.
Mind you, many cities, like my home city, unfortunately, doesnt have anything better locally other than going to a high-end roaster yourself and making your own coffee. But who wants that.
I pay for it (rather my business does) because its so damn convenient.
I also have a Sprint EVDO RevA card, but I tend to use the t-mobile at starbucks alot more because:
1. There's always one just right down the street.
2. The quality of the service is excellent, and its NEVER down.
3. It's much faster and lower latency than my sprint card.
All that being said, you're right in that its not a reasonable deal for casual users. But a business offering wifi at its stores is more complicated and expensive than you think, and there are tradeoffs at every point.
Ahh, but now you're providing a for-fee service.
It's got to be monitored, maintained, etc. You've got to have folks on standby to deal with it when it breaks.
You have to have engineers and software folks integrate the POS software with your wifi management software.
You've got to deal with software updates, hardware updates, hardware failures, line failures, attacks, spamming, and over-use.
The second you do anything but put it up and say 'enjoy it if it works', it becomes tremendously expensive.
But if you do put it up for free, you get saddled by freeloaders and people who abuse it, which makes it not work very well for the paying customers.
The problem is that its not as easy as people think. The second you make it a part of a business, the economics change tremendously.
As I've said elsewhere, there are tradeoffs.
In my experience, a large percentage (ie, the majority) of 'free' wifi spots are down or so slow to be useless a significant amount of time. This is a hassle, and costs me.
Whereas if I go to a starbucks, its always up (4+ years with 100% service for me), and the speed and latency are excellent (1.5mbps synchronous, good latency).
This quality guarantee is worth the $30/month for many people.
You're right in that it DOES suck for the casual user though. Not everyone makes their living off their computer and so can justify spending the money.
However, for people like myself and others, the T-Mobile/Starbucks thing has been a godsend the past 5 years.
Quality coffee, with high consistency, and a high-quality wifi connection anywhere in the country that you go. Hard to beat it, from a business user standpoint.
Agreed.
In addition, the quality of service of the wifi at Starbucks is excellent.
I've been to so many 'free' wifi spots where the wifi was down alot, or just so badly managed and/or congested and/or oversubscribed that it was useless.
I'd rather pay a few bucks a month and have something that 'just works' than have to roll the dice each time I sit down at a coffee shop.
I've looked at my traces and such while in there, and it doesnt appear to be a cellular network, nor does it have the ping times and latency associated with it.
In addition, T-Mobile has been present in my city's starbucks for years before any sort of cellular data service was available here (that could reach 1.5mbps at least).
Next time I'm over there I'll take another look, but I dont think this is correct.
It's possible they do this for some sites where they cant get dsl, but not generally. every place that has 'free' wifi, is a place where they put in high speed internet for their 'inventory' system, and the 'free wifi' piggybacks on that internet connection. in some cases, they use satelite for the inventory system ugh. I'm also fairly sure this isnt true. I've been to a large chunk of the major cities in the US, and have visited multiple starbucks in each one, and I've never seen a 'free' wifi at a Starbucks.
Although what you say sounds reasonable in theory, it doesnt work out that way in the real world.
Even people using multiple deathstar drives didnt see consistent failure times.
Very very few drive failures are because of defects. The deathstar notwithstanding. Even that though isnt relevant because no one in their right mind would use a drive like that for something important. I mean
Everyone that has released data and in my personal experience, hard drive failures are quite random and arbitrary. Even when a huge number of them is from the same manufacturer and model.
So please be careful what kind of mal-information you bring here, because someone might believe you and try to carry this bad information back to the workplace.
The risk level on RAID is quite simple. It's a ratio between expected drive failure rate and rebuild time.
There's a reason corporate systems still tend to build arrays out of 36, 75, or 150GB drives. It's because rebuild rates are reasonable (and also more spindles equals more speed). So if you ever see someone building a raid array out of 500GB consumer level drives, you know they are either:
1. Working with data that can take a much higher risk of multi-drive failure, since rebuild times are huge.
2. Working with multiple systems networked, and using replication between them to handle multi-drive failure risk.
3. Working with data availability needs that can stand going to tape if the whole array downs.
4. Idiots.
If your rebuild rate is more than an hour or so on very important data, then you use 10 or 6 or one of the higher raid levels.
So you plan what is the cost of losing a second drive while the first is rebuilding, and you design your storage around that.
Like everything else in IT, you pay for how much risk reduction you want.
In my experience, 2 drives failing in the same array close enough in time to lose the array is extraordinarily rare. However, sa user-error while trying to rebuild the array (ie, do it wrong and down the whole array, or pull the wrong drive) is fairly common.
If that happens, you go to your backups. If you cant afford the downtime by that, then you use a RAID level that you can afford and provides you the level of protection you need.
But RAID is not useless, and telling people that is just feeding bad information to people who dont know any better.
Then why is there an entire new dynamic language support infrastructure (ie, JSR292, invokedynamic) being built into Java 7?
My understanding is that JRuby and similar on the JVM use wrappers and conversions (basically a custom translation interface), whereas they are working on a more generically supported approach.
Some quick googling on jsr292 invokedynamic and java7 should give some more info.
A quick google myself shows some good info here.
Jim Hugunin has talked about some of the challenges of dynamic languages on the JVM compared to the DLR, which you can find with some googling.
PyProtocols and the Zope Interfaces look interesting, didnt even know there was such a thing out there.
I appreciate the information, and I'll look a bit more into it.
I saw some of your stuff after I wrote that response, but appreciate you posting the links anyway.
I'd like to see a more thorough analysis of some of this stuff by someone more qualified by me, using actual debuggers and signal analysis on the bus.
However, in the meantime, your tests are fairly straightforward (also glad Ostermann added some useful points about IHV added effects and whatnot), and match my personal experience.
This hasnt been an issue since SP2 came out on WinXP, and you know this.
He never (according to his website) investigated Vista. It's all second or third hand, or worse. All of this because Gutmann dared publish a careful and objective review of Vista's DRM. Except thats not what Gutmann did.
Gutmann read some very old white papers. He then built a mental model of what Vista might look like if people implemented it according to his personal interpretation of those white papers.
He then assumed that mental model to be true, and slammed this theoretical model that might be Vista on his interpretation of how MS might have implemented it.
But read his website closely. He clearly, at least according to anything I can find on the website, never once tried to actually see if any of his theories were correct.
Which is really strange, because thats the fundamental scientific method. Postulate a theory, then test your theory. Gutmann only ever did the first part. Everything Gutmann said has proved correct. Interestingly enough, no. There hasnt really been any refutation or support for any of Gutmann's claims, one way or another.
If you have a decent machine and good drivers, Vista clearly doesnt behave at the level of flakiness that Gutmann claimed it would, which seems to automatically invalidate some of his claims.
But the verdict is out, as far as I've seen. No one has ever attached hardware to the bus and tried to see if there are flip bits, or the kind of intrusive checking that Gutmann claimed would be in there.
It might be, it might not. No one that I have read has EVER tested it, including (as best I can tell from his publishings) Gutmann.
The silverlight plugin is also quite small (a couple of megs).
I agree, though for a little bit different reason.
Flex/Flash IS a much strong platform now, plus all of it except the Flash player is open-source (blazeds, amf, swf format, flex sdk, etc).
However, give a few years, and Silverlight/DLR/XAML will be quite the hot thing, technically. You'll always be strapped to MS using it, but aside from that, the underlying technology design is superior, IMO. Silverlight and the DLR are just way too young at this point.
But look at some of the comparisons:
Silverlight: XAML, Flex: MXML
MXML has quite a legacy background in movie/animation flash. It's evolved from there to a more typical event driven application. XAML was developed from the ground up for this and you can see it in how easy some things are to do in it. But XAML is also way young, and missing some things that just arent implemented in this version.
Silverlight: C#, any DLR language, Flex: ActionScript 3
Good lord, not even a comparison here. Just about any modern language is better than javascript/actionscript. Mind you, AS3 is a big leap over AS2 and previous. But still. It's javascript at the core, which is never good.
All that being said, we're opting for Flex on the front end and java/spring/hibernate on the back end. Mostly though because Flex is ready now, and platform independence and not being tied to MS is important to us.
Thats because its too new at the moment.
They did a quick release of Silverlight 1, which was pure XAML markup and javascript. Very light, very simple, and you could only do media-style apps on it really.
Silverlight 2 is in beta now, and is quite impressive technically, but very very young yet. XAML is quite nice, and being able to use C# or any DLR language in your silverlight code (as opposed to actionscript in flex/flash) is very very nice.
However, its very young, and wont have some traction for a while.
Flex/Flash isnt quite as sexy (technically) but has the advantage of alot more maturity, and less vendor dependency.