The most common one is the that a power failure results in a write hole, if the power fails between a data and parity write.
Note that this is a problem that only very very bad sysadmins would run into.
On any typical corporate system that has a RAID controller, you've got two levels of power protection:
1. UPS for the room or the rack or the box. 2. Battery backed write cache on the controller.
This means that even if BOTH of your redundant PSUs fail simultaneously, as long as they dont physically torch the controller, then the controller will have the unwritten cache stored.
Re:RAID5 is stupid, RAID 10 or no RAID
on
What NAS To Buy?
·
· Score: 1
You should go back and read the article you linked about RAID10.
I'll summarize.
RAID10 is a stripe of mirrors. So for example, if you have six discs, A-F:
AB - mirrored CD - mirrored EF - mirrored
Then you have those three 'mirrored' drives set as a non-redundant stripe.
The pictures on the link you included are very clear.
So given this, think about what happens when you lose both sides of one mirror.
It's precisely the same as losing any one disc from a non-redundant stripe set.
RAID10 will fail if you lose any two sides of a mirror.
RAID01 isnt really used in the real world. A mirror of two stripe sets? What the hell would be the point?
Re:RAID5 is stupid, RAID 10 or no RAID
on
What NAS To Buy?
·
· Score: 1
So you're talking about a temporary known issue in a single product 4+ generations ago, and from there extrapolating that they all do that and ie suck?
PERC is up to gen 6 now, and I've been personally involved in managing server rooms since the PERC3 days, and have never seen nor heard of this problem.
Re:RAID5 is stupid, RAID 10 or no RAID
on
What NAS To Buy?
·
· Score: 1
This is absolutely and positively not correct, for the business class stuff from Dell.
The PERC controllers in the PowerEdge servers monitor constantly through a couple mechanisms:
1. If a drive fails, it'll bring the array into degraded mode instantly, and start automatically rebuilding the array from the hot spare if thats how you have it configured.
2. Any SMART errors will get reported to the Server Manager and will email you or whatever kind of alert configuration you have. We've had these warn us ahead of drive failures many many times over the years.
3. Patrol Reads. You can configure the PERC controllers to continuously do read/write tests on random locations throughout the array. It's a way of proactively finding bad disks before they fail completely.
HP = Compaq, and has been for many years.
Re:RAID5 is stupid, RAID 10 or no RAID
on
What NAS To Buy?
·
· Score: 1
In that case, your array is just fine. It's not accessible, but you havent lost any data.
So you either drop another controller you have in stock, or have one fedex'd out the next am.
Most raid controllers write the raid configuration on both the drives and the raid battery backed cache, so its not like you'll lose the data.
Re:RAID5 is stupid, RAID 10 or no RAID
on
What NAS To Buy?
·
· Score: 1
With Raid5.. two drives fail and your done.
Only if they fail at the same time. It only needs enough time after one failure to rebuild the drive before you have full redundancy.
Unless you buy every drive at a different time from a different manufacturer, chances are under the same wear conditions, two will fail around the same time.
I wish people would stop spreading this meme. It's not true in the real world (one or two historical special cases notwithstanding, ie deathstar drives, but then again, those should never have been used in a raid array).
It's one of those arguments that sounds logical and makes sense, but doesnt really work that way in the real world.
Raid10 can withstand a much larger failure
Kind of, but not really. If you lose two of the WRONG drives, you're just as dead in the water as with losing two in a RAID5 (ie, losing both sides of one mirrored pair). The flip side is that if you lose all the RIGHT drives, you can lose half your drives and still keep chugging.
If being able to withstand multiple-disk failure is the most important thing (and especially more important than raw write performance) then use a multiple-disk-parity solution (ie, RAID6 and so forth). These lets you lose ANY 2+ (depending on the number of parity drives) drives and keep going.
and you also get some serious performance++.
Yes and no.
Read performance is mostly based on the number of spindles you have dishing content. With RAID5, its N-1 spindles that can deliver any given file. With RAID10 its N. Not a material difference there.
On writes is where it can be a noticeable difference, due to the parity calculation. But it really isnt a huge problem in most loading scenarios.
Ignoring that there is a much bigger hole in IE that the Apple bug makes a tiny bit easier to trigger I would disagree as to which one is the bigger hole.
Anything that lets arbitrary attackers write arbitrary files to protected locations on the local system is worse than IE loading DLLs from known locations.
Why? Simply because if you're able to write files to the computer (outside of cookies and temp) just by having someone visit a website then you've largely owned the computer at that point.
The only thing that restricts the scope of the apple bug is that it only writes to the desktop (which is a stupid auto-download location).
IE's issue with loading DLL's IS stupid, but by itself is a complete non-issue.
Whereas Apple's bug by itself is still problematic.
To be fair though, MS IE has had a very turbulent history when it comes to drive-by file dropping onto the local system. But in this specific case, I would disagree with your statement.
Apple finds serious bugs in Ruby. They tell the Ruby developers. Ruby developers issue patches.
MS finds a bug in Safari. They tell the Apple developers. Apple developers say they wont patch it soon. MS then tells everyone not to use Safari until its fixed. You're right in that its not the same situation, but when you put all the facts in, rather than trying to cast a one-sided light on the situation, its alot clearer.
PS, the funky quoting style is slashdot's recent bugfest that ignores hard returns thats cropped up lately.
A more relevant comparison would be.NET to Python. I think you'll find that.NET fares quite well in that comparison, at least according to the information at Secunia.
I think you need to look at the disclosure histories yourself.
Assuming all things are equal,.NET has by far the best record. Python in the middle (by raw count), and Java at the end.
Mind you, Python has two 'own your system' unpatched vulnerabilities right now, that are between 6 and 9 months old and still unpatched. They could be less serious than secunia makes them out to be, however, I'm not familiar enough with them to say off the top of my head.
I'm not going to do it again here, but I also looked at and linked to the secunia listings of.NET and Java in a post just above here..NET has an excellent record. Java less so, but still not terrible.
You're exaggerating the risk of the Java JVM and particularly.NET quite a bit.
If you look at the security hole history of.NET 1.1,.NET 2.0, and.NET 3.0, you'll notice an almost perfect history.
The only true easy own your box was the JPEG parsing vuln that affected a ton of MS products, and that hit.NET as well, due to shared code/modules.
The JVM has been less close to perfect, but its not too bad. You can read about them for JRE 1.4, JRE 1.5/5, and JRE 1.6/6.
I would also say that its not an apples to apples comparison. Most of the vulns in.NET and Java have been not in the core language itself, but in the web-applet piece, or in image handling or similar parts of the libraries built in. These are much larger than the built-in libraries that Python ships with.
I'm not trying to start an argument of who has the most possible libraries, including 3rd party, but just pointing out that the default shipment of Java and.NET comes with alot more 'stuff', which widens the attack surface area.
I'm not sure that I agree that the perl thing is more serious.
This is an integer overflow bug, which allows someone to write arbitrary code to arbitrary points in memory.
It's just a short matter of time before that is 'weaponized' as TFA puts it, and shellcode inserts are achieved through this.
At this point, this is a remote, instantly-own-your-box without requiring any user access to it at all.
The suidperl issue is a priv escalation, which means you have to have an unpriv'd account in the first place. The Ruby one doesnt require that.
Mind you, the ruby holes havent been 'weaponized' yet publicly, but there are a lot of black hats with alot of experience using integer overflow bugs into own your box exploits.
It's easy to hit for a developer, running this on their desktop:
Eclipse with plugins (~1GB) Oracle Server or Express (~1GB) Tomcat/JBoss/etc running your app (~1GB)
Now, all of those things dont require a full gigabyte of space every second, but its common to set them up that way so that you dont ever run out of memory when working on your apps.
Similarly for an MS focused developer with VS2008, SQL Server 2005 developer, and IIS.
Or any subset of that plus a VM running something else for testing.
From what I understand, there are two basic ways: Drive by downloads and host programs that carry spyware with their installation. This is incorrect, at least in my experience. The vast majority of malware installs I see are from people installing it explicitly. They think its a porn codec or a cool video player, or a cracked copy of Office or something.
So nothing to do with software, just pure social engineering.
Drive by downloads under Windows are installed thanks to Internet Explorer bugs. IE is capable of installing operating system updates and so it automatically has the access needed to do so.(*) Safari has no special operating system privileges and so it cannot install software on its own without user intervention. This is completely inaccurate and shows a lack of understanding of operating system behavior.
IE isnt magic, and doesnt have magical properties. It is simply an executable that runs under some account.
If you run it under an account with local admin privileges, then it has local admin privileges. If you run it under an account with no admin privileges, then it has no admin privileges. If you run it with a 'drop my rights' style tweak, then it has exactly those rights.
The bad mix is people running IE under an account with local admin privs. Then the various patched and unpatched vulns expose the user to random software installation for the joy of browsing to a webpage.
And on Win I can't setup my kids' account as a non-Admin: Alpha Centauri would neither install nor run from a non-Admin account. I think what you mean to say is that you most certainly can setup your kids on windows as non-admin.
It just happens that one piece of software they like is so horrendously designed and implemented that it requires admin to run.
What did the Alpha Centauri developers say about this when you approached them about fixing their software?
If so, there's going to be lots of problems, as other ODF-reading software already does this.
ODF is not a big enough spec yet to encompass everything anyone would want in an office document. Therefore to be competitive, any office document editor would have to add extensions.
Even if thats true (which I'm not sure of), who cares?
Everything on the planet can read NTFS. It's not like that leaves us locked into an inaccessible format, even if MS disappeared off the face of the planet today.
or are you not good enough ? if so how are you qualified to comment and why should we listen to you ? Wow. I mean... wow.
Way to completely show off the precisely wrong attitude for a successful software franchise.
You've got it precisely backwards.
The SINGLE MOST IMPORTANT PEOPLE to listen to are the non-technical users. Why? Because they're 99% of the population.
If the only people you ever listen to are people capable of writing the software, then you're forever dooming your product to only being useful to those same people.
If thats your goal, then fine. A shiny tool for devs. Lots of good small open source tools are like that.
For people who are a bit bigger the Criterion Plus is nice.
It's a 'big and tall' chair (basically a massively oversized and overstuffed version of the regular criterion task chair) from Steelcase, but for someone like me, who is just a little bit big and tall, its like a couch chair.
It's insanely sturdy and tough, thick padded armrests, hugely overstuffed seat with high density foam, and a tall back.
The seat and armrests are wide enough that I can really slouch and spread out if I want, and the armrests are farther out than most chairs.
Would be too much for a skinny 5'8" person, but for someone who is 6'2" and not thin, its a great chair.
Wheels - being able to push the chair back when you move away from the desk (otherwise you will probably just tear the carpet). Note that this is why you put the big flat plastic thing down on the carpet under your chair (its late and I cant remember what they're called right now).
In fact, most office buildings that you lease have it in their lease agreement, because a big guy sitting in a chair in the same place for 6 hours a day will destroy the carpet without one.
On the other side, Thunderbird default quick find-as-you-type search is about 1000 times faster than outlook. Even with Outlook 2007? Mine is instant on a 1.2GB exchange mailbox (running in cached exchange mode).
But it wasnt very fast at all with Outlook 2003 unless you installed Lookout.
Overall, Thunderbird has always been more responsive to me and this is always what I recommend. For me, if all they need is raw email and they can grok it being stored on the server, I do recommend Thunderbird and IMAP. But nearly all business customers I go to Outlook/Exchange first. It's expensive and complicated but man is it empowering, especially shared in your org.
I'm not really sure where you are taking your stats, but I remember seeing some market share analysis of mail servers a few months ago and Exchange was still second to IBM Lotus Notes (mind you, stats change often). I personally could count on my fingers the number of companies that have it. Yeah, Lotus has a handful of very large customer bases in the US which really grows those numbers. Which is weird because for an email client, thats like the worst user experience I've ever seen.
It could be part of the world maybe... but the vast vast majority of non-technology small and medium businesses I deal with use Exchange. The small ones via SBS, and the bigger ones with regular exchange.
This would include only companies that have an IT department and not the ones working with outsiders. There are several decent outsource Exchange hosting for as little as $7 per month per user for 1GB of space, up to $25-30 a month per user for 2GB of space and a better service level. Neither require an IT department... at most may require a consultant like myself to do initial setup, but any reasonably technical in-house person could do that.
I'm not sure what the 'working with outsiders' means. My business is all working with outsiders (software and IT consulting) and Exchange doesnt provide a barrier to me.
In fact, I have clients send me exchange calendar requests all the time (on different domains & exchange servers) on accident, because they just dont know that it might not work. And it works great.
Another point regarding this, since Outlook doesn't really play fair with IMAP standards (and Thunderbird does), I am switching most of these clients to Thunderbird. The only one that is reluctant is the boss and he is still happy using Outlook 97 from an old copy. Agreed.
Outlook is terrible on IMAP.
Outlook is only worthwhile if you're using Exchange, or for small clients who POP but like the calendars and tasks.
The biggest problem (IME) with Outlook against IMAP is that it doesnt handle deleted messages right.
Just crosses them out and leaves them in your inbox. Even the option to hide them doesnt help as you have to delete them eventually.
There is probably a way to write a vba script to handle deletes to do it like every other mail client on the planet and move it to the deleted items folder, but I never did get that working.
Slashdot Mirror
The most common one is the that a power failure results in a write hole, if the power fails between a data and parity write.
Note that this is a problem that only very very bad sysadmins would run into.
On any typical corporate system that has a RAID controller, you've got two levels of power protection:
1. UPS for the room or the rack or the box.
2. Battery backed write cache on the controller.
This means that even if BOTH of your redundant PSUs fail simultaneously, as long as they dont physically torch the controller, then the controller will have the unwritten cache stored.
You should go back and read the article you linked about RAID10.
I'll summarize.
RAID10 is a stripe of mirrors. So for example, if you have six discs, A-F:
AB - mirrored
CD - mirrored
EF - mirrored
Then you have those three 'mirrored' drives set as a non-redundant stripe.
The pictures on the link you included are very clear.
So given this, think about what happens when you lose both sides of one mirror.
It's precisely the same as losing any one disc from a non-redundant stripe set.
RAID10 will fail if you lose any two sides of a mirror.
RAID01 isnt really used in the real world. A mirror of two stripe sets? What the hell would be the point?
So you're talking about a temporary known issue in a single product 4+ generations ago, and from there extrapolating that they all do that and ie suck?
PERC is up to gen 6 now, and I've been personally involved in managing server rooms since the PERC3 days, and have never seen nor heard of this problem.
This is absolutely and positively not correct, for the business class stuff from Dell.
The PERC controllers in the PowerEdge servers monitor constantly through a couple mechanisms:
1. If a drive fails, it'll bring the array into degraded mode instantly, and start automatically rebuilding the array from the hot spare if thats how you have it configured.
2. Any SMART errors will get reported to the Server Manager and will email you or whatever kind of alert configuration you have. We've had these warn us ahead of drive failures many many times over the years.
3. Patrol Reads. You can configure the PERC controllers to continuously do read/write tests on random locations throughout the array. It's a way of proactively finding bad disks before they fail completely.
HP = Compaq, and has been for many years.
In that case, your array is just fine. It's not accessible, but you havent lost any data.
So you either drop another controller you have in stock, or have one fedex'd out the next am.
Most raid controllers write the raid configuration on both the drives and the raid battery backed cache, so its not like you'll lose the data.
With Raid5 .. two drives fail and your done.
Only if they fail at the same time. It only needs enough time after one failure to rebuild the drive before you have full redundancy.
Unless you buy every drive at a different time from a different manufacturer, chances are under the same wear conditions, two will fail around the same time.
I wish people would stop spreading this meme. It's not true in the real world (one or two historical special cases notwithstanding, ie deathstar drives, but then again, those should never have been used in a raid array).
It's one of those arguments that sounds logical and makes sense, but doesnt really work that way in the real world.
Raid10 can withstand a much larger failure
Kind of, but not really. If you lose two of the WRONG drives, you're just as dead in the water as with losing two in a RAID5 (ie, losing both sides of one mirrored pair). The flip side is that if you lose all the RIGHT drives, you can lose half your drives and still keep chugging.
If being able to withstand multiple-disk failure is the most important thing (and especially more important than raw write performance) then use a multiple-disk-parity solution (ie, RAID6 and so forth). These lets you lose ANY 2+ (depending on the number of parity drives) drives and keep going.
and you also get some serious performance++.
Yes and no.
Read performance is mostly based on the number of spindles you have dishing content. With RAID5, its N-1 spindles that can deliver any given file. With RAID10 its N. Not a material difference there.
On writes is where it can be a noticeable difference, due to the parity calculation. But it really isnt a huge problem in most loading scenarios.
Anything that lets arbitrary attackers write arbitrary files to protected locations on the local system is worse than IE loading DLLs from known locations.
Why? Simply because if you're able to write files to the computer (outside of cookies and temp) just by having someone visit a website then you've largely owned the computer at that point.
The only thing that restricts the scope of the apple bug is that it only writes to the desktop (which is a stupid auto-download location).
IE's issue with loading DLL's IS stupid, but by itself is a complete non-issue.
Whereas Apple's bug by itself is still problematic.
To be fair though, MS IE has had a very turbulent history when it comes to drive-by file dropping onto the local system. But in this specific case, I would disagree with your statement.
Nice job with the random hyperbole there.
Lets report on this more accurately:
Apple finds serious bugs in Ruby. They tell the Ruby developers. Ruby developers issue patches. MS finds a bug in Safari. They tell the Apple developers. Apple developers say they wont patch it soon. MS then tells everyone not to use Safari until its fixed. You're right in that its not the same situation, but when you put all the facts in, rather than trying to cast a one-sided light on the situation, its alot clearer.PS, the funky quoting style is slashdot's recent bugfest that ignores hard returns thats cropped up lately.
Comparing Python to Windows is a bit silly.
A more relevant comparison would be .NET to Python. I think you'll find that .NET fares quite well in that comparison, at least according to the information at Secunia.
I think you need to look at the disclosure histories yourself.
Assuming all things are equal, .NET has by far the best record. Python in the middle (by raw count), and Java at the end.
Mind you, Python has two 'own your system' unpatched vulnerabilities right now, that are between 6 and 9 months old and still unpatched. They could be less serious than secunia makes them out to be, however, I'm not familiar enough with them to say off the top of my head.
Python 2.3.x
Python 2.4.x
Python 2.5.x
Python 2.6.x
I'm not going to do it again here, but I also looked at and linked to the secunia listings of .NET and Java in a post just above here. .NET has an excellent record. Java less so, but still not terrible.
You're exaggerating the risk of the Java JVM and particularly .NET quite a bit.
If you look at the security hole history of .NET 1.1, .NET 2.0, and .NET 3.0, you'll notice an almost perfect history.
The only true easy own your box was the JPEG parsing vuln that affected a ton of MS products, and that hit .NET as well, due to shared code/modules.
The JVM has been less close to perfect, but its not too bad. You can read about them for JRE 1.4, JRE 1.5/5, and JRE 1.6/6.
I would also say that its not an apples to apples comparison. Most of the vulns in .NET and Java have been not in the core language itself, but in the web-applet piece, or in image handling or similar parts of the libraries built in. These are much larger than the built-in libraries that Python ships with.
I'm not trying to start an argument of who has the most possible libraries, including 3rd party, but just pointing out that the default shipment of Java and .NET comes with alot more 'stuff', which widens the attack surface area.
I'm not sure that I agree that the perl thing is more serious.
This is an integer overflow bug, which allows someone to write arbitrary code to arbitrary points in memory.
It's just a short matter of time before that is 'weaponized' as TFA puts it, and shellcode inserts are achieved through this.
At this point, this is a remote, instantly-own-your-box without requiring any user access to it at all.
The suidperl issue is a priv escalation, which means you have to have an unpriv'd account in the first place. The Ruby one doesnt require that.
Mind you, the ruby holes havent been 'weaponized' yet publicly, but there are a lot of black hats with alot of experience using integer overflow bugs into own your box exploits.
It's easy to hit for a developer, running this on their desktop:
Eclipse with plugins (~1GB)
Oracle Server or Express (~1GB)
Tomcat/JBoss/etc running your app (~1GB)
Now, all of those things dont require a full gigabyte of space every second, but its common to set them up that way so that you dont ever run out of memory when working on your apps.
Similarly for an MS focused developer with VS2008, SQL Server 2005 developer, and IIS.
Or any subset of that plus a VM running something else for testing.
So nothing to do with software, just pure social engineering.
Drive by downloads under Windows are installed thanks to Internet Explorer bugs. IE is capable of installing operating system updates and so it automatically has the access needed to do so.(*) Safari has no special operating system privileges and so it cannot install software on its own without user intervention. This is completely inaccurate and shows a lack of understanding of operating system behavior.IE isnt magic, and doesnt have magical properties. It is simply an executable that runs under some account.
If you run it under an account with local admin privileges, then it has local admin privileges. If you run it under an account with no admin privileges, then it has no admin privileges. If you run it with a 'drop my rights' style tweak, then it has exactly those rights.
The bad mix is people running IE under an account with local admin privs. Then the various patched and unpatched vulns expose the user to random software installation for the joy of browsing to a webpage.
It just happens that one piece of software they like is so horrendously designed and implemented that it requires admin to run.
What did the Alpha Centauri developers say about this when you approached them about fixing their software?
Define non-compliant.
Is it non-compliant if its just got extensions?
If so, there's going to be lots of problems, as other ODF-reading software already does this.
ODF is not a big enough spec yet to encompass everything anyone would want in an office document. Therefore to be competitive, any office document editor would have to add extensions.
Right click on the Ribbon, choose 'Minimize the Ribbon'.
Makes it work like the start menu in 'auto-hide' mode. Will solve your real estate problems.
Note that SVN and the like will work quite well with the Office 2007 formats and ODF from MS Office.
Since both formats are just XML files zipped, it would probably only require a very simple plugin for SVN to handle these very gracefully.
You'll still just get a big ball of base64 or equiv when you've got embedded images, but thats a known factor.
Even if thats true (which I'm not sure of), who cares?
Everything on the planet can read NTFS. It's not like that leaves us locked into an inaccessible format, even if MS disappeared off the face of the planet today.
Way to completely show off the precisely wrong attitude for a successful software franchise.
You've got it precisely backwards.
The SINGLE MOST IMPORTANT PEOPLE to listen to are the non-technical users. Why? Because they're 99% of the population.
If the only people you ever listen to are people capable of writing the software, then you're forever dooming your product to only being useful to those same people.
If thats your goal, then fine. A shiny tool for devs. Lots of good small open source tools are like that.
But I dont think Open Office fits that bill.
Those are nice.
For people who are a bit bigger the Criterion Plus is nice.
It's a 'big and tall' chair (basically a massively oversized and overstuffed version of the regular criterion task chair) from Steelcase, but for someone like me, who is just a little bit big and tall, its like a couch chair.
It's insanely sturdy and tough, thick padded armrests, hugely overstuffed seat with high density foam, and a tall back.
The seat and armrests are wide enough that I can really slouch and spread out if I want, and the armrests are farther out than most chairs.
Would be too much for a skinny 5'8" person, but for someone who is 6'2" and not thin, its a great chair.
In fact, most office buildings that you lease have it in their lease agreement, because a big guy sitting in a chair in the same place for 6 hours a day will destroy the carpet without one.
But it wasnt very fast at all with Outlook 2003 unless you installed Lookout. Overall, Thunderbird has always been more responsive to me and this is always what I recommend. For me, if all they need is raw email and they can grok it being stored on the server, I do recommend Thunderbird and IMAP. But nearly all business customers I go to Outlook/Exchange first. It's expensive and complicated but man is it empowering, especially shared in your org. I'm not really sure where you are taking your stats, but I remember seeing some market share analysis of mail servers a few months ago and Exchange was still second to IBM Lotus Notes (mind you, stats change often). I personally could count on my fingers the number of companies that have it. Yeah, Lotus has a handful of very large customer bases in the US which really grows those numbers. Which is weird because for an email client, thats like the worst user experience I've ever seen.
It could be part of the world maybe
I'm not sure what the 'working with outsiders' means. My business is all working with outsiders (software and IT consulting) and Exchange doesnt provide a barrier to me.
In fact, I have clients send me exchange calendar requests all the time (on different domains & exchange servers) on accident, because they just dont know that it might not work. And it works great. Another point regarding this, since Outlook doesn't really play fair with IMAP standards (and Thunderbird does), I am switching most of these clients to Thunderbird. The only one that is reluctant is the boss and he is still happy using Outlook 97 from an old copy. Agreed.
Outlook is terrible on IMAP.
Outlook is only worthwhile if you're using Exchange, or for small clients who POP but like the calendars and tasks.
The biggest problem (IME) with Outlook against IMAP is that it doesnt handle deleted messages right.
Just crosses them out and leaves them in your inbox. Even the option to hide them doesnt help as you have to delete them eventually.
There is probably a way to write a vba script to handle deletes to do it like every other mail client on the planet and move it to the deleted items folder, but I never did get that working.
Thank you for posting this.
I've been complaining about this issue for years, and did not know this was possible.
Thank you so much for posting this!