I don't have to prove that my imaginary die is perfectly balanced.
You would have to prove that your perfectly damaged die with independent trials is comparable to the OP's situation; which it's not.
The proposition you are making applies to only a specific kind of statistical argument.
In the real world there are plenty situations where a previous outcome influences the next.
I will also add that in the next 50 days I am far more likely to get attacked by a tiger than in the past 50 days.... because I didn't get attacked by a tiger in the last 50 days.
That only works if you cheated and already observed the last 50 days.
If I asked you to go deep into tiger country, unprotected, disguised as the Tiger's prey, for 100 days, Or for 50 days.
Than with 100 days, your chances of getting attacked are larger.
Now; if you already observed 50 days, then you already know you weren't, so your chances of getting attacked during the next 50 are not necessarily any higher.
But they may be higher; since you will be wandering into yet further new territory, and there will be yet more time and more tracks left, for the tigers to find.
They may catch on to a trail you left during the last 50 days, so they may actually be more likely to find you on day 51, than on day 1.
The assumption that Tiger attacks are entirely random and not related to your previous days spent in places at risk, is not a reasonable assumption.
I would want to drive manually most of the time, but it would be great to be able to have a built-in designated driver for trips to the bar:)
As long as you have the physical capability of overriding the computer and driving it yourself -- then you have a possibility of having control of the vehicle; therefore, it is illegal DUI, even if the computer is driving it at the moment: the fact that you can override it at will makes it still an offense.
I suspect the law won't be revised to allow you to be drunk: until an inspection proves that a drunk driver cannot in a moment of weak judgement override their computerized vehicle.
Hey, I don't care if someone wants a self-driving car.
As long as it doesn't hinder, re prohibit MY ability to buy and drive manual cars on the road.
I feel that's OKAY, but once self-driving cars are proven; the sale of new vehicles that can be manually operated should be restricted to specialty vehicles custom made or produced and sold in small quantities, and for commercial buyers that have special requirements.
And liability insurance companies should be allowed to provide discounts for the reduced risk to operators;
with the costs shifted to operators of manual cars, additional licensing qualifications, and blackboxes required, to save lives, and allow society to benefit from the elimination of auto accidents that smart cars should result in.
it would be absurd for me to believe that I would likely get a one the next 100 time. No, for the next 100 rolls I should expect to get about 16 or 17 ones.
Dice rolls have an important statistical property that doesn't exist in this situation.
Dice rolls are independent. One dice roll does not affect the outcome of another dice roll.
That's not automatically true: in general it is not true, and you have to prove that it is true.
Most dice won't be perfectly balanced.
It's possible that 250K miles is a magic number: you kept your car the whole time.
Some mechanical defect exists in a large number of cars that cause them to break down on average in between 250K and 300K miles; often causing an accident.
In that case... after having driven 250K miles: your next 50K milles are potentially more dangerous than your last 50K miles.
That isn't how statistics work. The one accident in 250k is an aggregate of a lot of data. It doesn't mean that it is expected that after you get to 250k and beyond you're more likely to get into an accident.
If you have driven 300K miles, you are more likely to have gotten into at least one accident having driven that many miles than if you had driven 250K miles.
Arguably; your probability of an accident per mile driven is nonuniform and nonindependent; miles driven in a parking lot have a lower weight --- if your driveway is longer than other people's, then you might drive more miles with lower expectation of an accident.
When you are experienced and have driven less than 10000 miles in your life; I would argue, that you are more likely to get into an accident in those 10000 miles than the next 10000 miles.
I would also argue that miles driven in heavy traffic or bad weather are miles that you are more likely to have an accident in --- you could have driven 100K miles in ideal conditions, and gotten no accident, by driving less safely than the guy who drove 50K miles in horrid conditions, but got into an accident.
I do not recommend accident-free miles as a metric of driver quality.
I would argue, that a large number of miles driven suggests you are a better driver than the bottom 1/3rd,
because of the road experience you got, having driven all those miles.
The more hazardous the conditions you drove more miles under with no accidents,
and the more of those miles that were driven on congested high-speed roadways and bumper-to-bumper traffic,
and other miles where accidents are most likely to occur --
the better a driver you are.
They could have just listed the fatality rates of the different countries; or provided a color-coded list.
For it to be useful as a map; it should be more granular, than merely painting every country the same color.....
it should show fatality rates for states, provinces, counties, cities, and individual streets.
Now that would make sense as a map.
Then I'd have to say you probably haven't worked much administering Windows networks of any size or complexity. That wasn't even true under Windows NT 3.51 when I worked with it in 1996.
What exactly are you doubting?
You have a domain controller blue screen.... happens all the time. Take a few minutes, maybe an hour for someone to get to physically reboot it; by the time it comes back up a JRNL_WRAP condition has occured, and SYSVOL replication is now officially borked, and group policy cannot be distributed.
Plus as luck would have it... there is frequently one form of corruption or another in the AD database or SYSVOL, so broken versions of the group policy files or AD database get spread around; because NTFS, has such poor resiliency and of the 100 sites in the domain each with their own domain controller.... one of them is bound to eventually have one form of data corruption or another....
Yep, and it works really, really well. Well enough that there are Windows machines sitting on the platforms of remote train terminals for years quietly doing their jobs without ever being cracked
If group policy is even used on those; then it's in spite of group policy -- not because of group policy.
Linux remote terminals would tend to have a self-contained configuration distributed from a central source, and redistributed to machines only when required.
Windows' reliance on domain controllers' continuous correct operation makes things fragile.
One of the top ten issues that Windows servers frequently develop is failure of replication, and in particular --- failure to apply or properly distribute group policy updates.
A rsync of some flatfiles is infinitely more resilient and reliable than what Windows does.
I might note that, if it weren't for the fact that the folks in Sales are addicted to their Outlook features, the entire rest of the company would be delighted to be rid of both Outlook and Exchange.
Seriously.... if that's the case; I would consider running a split-organization with Exchange for the Sales folks, and Postfix; Cyrus, or Zimbra for everyone else.
It is indeed possible to do that, even if all users have the same internet domain -- seeing that exchange has such a concept as an internal relay domain, where unknown recipients get pushed out to another server of your choosing.
def get_legislator_vote(Representative me, Bill b)
if me.received_campaign_contributions_from_supporter?
if me.can_spin_to_constituents(b, b.supporters)
return:Yes
else
return:Present
end
end
if me.received_campaign_contributions_from_opponent?
if me.can_spin_to_constituents(b)
me.add_actions(:acquire_favors, b.supporters)
return:No
else
me.add_actions(:cash_in_favor_to_get_no_votes)
return:Dont_Vote
end
end
if b.bill_gives_government_more_power? && me.can_spin_to_constituents(b)
return:Yes
end
if b.bill_restricts_government_power?
return:No
end
if b.cnn_says_bill_addresses_current_crisis?
return:Yes
else if b.cnn_says_bill_will_cause_apocalypse?
return:No
end
if me.party == b.billsponsor.party
if me.party == currentpresident.party
if currentpresident.opposes? b
return:No
else if currentpresident.supports?(b)
if me.canSpinToConstituents?(b)
return:Yes
else b.amendable?:Amend
else:Present
end
else return:Present
end
else
if currentpresident.supports
return:No
lse if currentpresident.opposes? return:Yes
else return:Present
end
else
return:No;
end
end
If you have got 50,000 messages in your Inbox, you will probably experience some bad performance, even with Gmail
Blah.... bad IMAP performance, as in your mail client will begin to have trouble. Of course... the Gmail web interface is marvelous, and you can probably have half a million messages in your inbox, as long as you don't want an IMAP software program to index that and download the headers of all those messages.
About half of the folks in my group use Thunderbird with the Exchange server, and the crashes and other problems hit randomly - sometimes it's 1/2 Outlook and 1/2 Thunderbird users whose mail is not working because the remote server is having a bad hair day.
I can understand Thunderbird users having issues. Exchange's implementations of POP and IMAP are poor and should be avoided, plus the default throttling policies may not be adequate to stop some runaway IMAP clients causing serverwide issues. I can't accept "remote server is having a bad hair day"; either it was misconfigured, sized with too few resources (Sufficient IOPS at low-latency, RAM, Memory, CPU, Disk), being used in a way above or outside the scale that Exchange is designed for, not being adequately protected against abusers or misbehaving clients, or there is a hardware issue, or problem with a service it's dependant on such as AD. Bugs in Exchange are possible as well, but if experienced, the ones that really matter show up in the logs as more than "having a bad hair day".
I'll accept that a 20GB remote inbox might have indexing & searching lags, if there isn't a fulltext index at the other end, but that's a different issue.
The PST/OST file format on the client is prone to corruption at those sizes, which means a big amount of downtime for the end user to redownload 20GB of data.
The bigger a mailbox gets, the more storage I/O capacity is required simply for accessing that mailbox.
Which means, that the entire server will run out of precious expensive IO resources with increasing speed, the bigger your users' mailboxes get.
Although.... to some limited extent, you can counter this by having extra RAM on your exchange server, above and beyond Microsoft's standard recommendations of ~50MB per mailbox, for extra Gigabytes of active mailbox database storage you use, by multiplying that by a factor, and requiring 200gb of RAM for the Exchange database server, instead of 32gb.
It also means that you can essentially have 10 users per database until you reach Microsoft's recommended database size maximum of 200GB; in the event that you have a much larger database, the delay required to repair or restore from backup can easily extend downtime to an unacceptable duration.
If your users want to keep their old mail, use an archiving system that was designed with an architecture suitable for operating at that scale.
New-MailboxRepairRequests and various other operations for large mailboxes result in unacceptable downtime for the user, when these are supposed to be routine, mostly transparent activities.
Bringing a system to conditions where that system cannot be made to recover gracefully from occassional errors or problems is a bad situation to have.
I'll also note that my GMail account probably has 30,000 emails in it right now.
It's not fair at all to compare Gmail to Exchange. The architecture is completely different.
Gmail's infrastructure is a much more scaleable system, they have loads of cheap storage in a proprietary magic Google clustering system that few can get close to matching ---- however, it's also managed by someone else, AND it's distant -- so if you need your mail fast, Gmail's not necessarily a good idea.
On the other hand, if you need large attachments, Gmail's probably a great idea.
Exchange's selling point is not that it's a scaleable system. It's that people are used to Outlook and like its features.
If you have got 50,000 messages in your Inbox, you will probably experience some bad performance, even with Gmail
And the note about inboxes - in our business we deal with a LOT of email that is several megabytes per message.
This is likely to weigh heavily on Exchange, best practices are to reject them -- to have a reliable system; remember E-mail is a store and forward system.
Messages that are more than a meg or so are a potentia
All? I guess there's no point in installing all those updates then.
Sure... install EMET, enable all the mitigations, rubber stamp the latest Windows 7 release; make sure your users don't install Java or Flash (ban the plugins), and you're pretty much as good to go, as if you installed every update --- your users will still manage to get malware on there, and not updating doesn't make it more likely or less likely.
For bonus points, add an AV or application whitelisting solution ala Bit9.
The antivirus has about as much chance catching a malicious file as a coin flip though.
Just a fact of life -- AVs used to be better, but have become THAT ineffective.
That's easy enough to say for someone who hasn't had to deal with 10 billion years of other lifeforms and their bullshit.
I don't care about that shit. I'd just ignore it all and play video games for a few eternities,
after cashing in the dollars in interest and stock dividends earned over my first 200 years of immortality.
What I wouldn't want to do is live forever but still age, live forever as an old person, or have the risk of
being injured in pain, hungry, trapped, or disabled and still live forever -- but in some horrible predicament; or be a person that has to put up with the ailments of old age or sickness.
I would rather be an immortal 6 or 7 year old girl who is uninjurable, invincible and can never feel pain or discomfort, than live forever as a 60, 70, or 80 year old, who society allows to do about whatever they want.
In other words: cryonics on old people is insane; unless we can stop aging.
It makes more sense for the young instead, heh....
If that happens, I wonder if they'll be able to keep their employees around, or if all the good ones will just leave.
If they'll bail out just because of some price swings of publicly traded stock, or they were just there because of insanely high valuations, then they are not very good employees.
You don't work for a company based on its stock price.
You invest in a company's shares based on its stock price being below what you believe the shares will be worth; or what you believe those shares will eventually be worth relative to the market and relative to your opportunity costs in regards to other options for investing your cash.
Yes, that problem of allowing everyone in a company to have a computer on their desktop.
No.... the problem is the Windows registry; which without group policy makes distribution of standard settings difficult.
Group policy is nothing more than a system for deploying canned registry hacks to specified computers, based on which checkboxes someoned ticked and saved to a GPO file.
P.S. Free speech or making an offensive movie are parole violations. Or at least..... well, other things that would otherwise be overlooked become parole violations, after the exercise of said free speech.
You don't like someone's political views? Get them arrested on same fake charge; get them parole.
But if they dare speak out again against our overlords, you can be sure it will be a parole violation
Yeah... after spending 24 hours on hold, then waiting another 24 hours for a callback, and getting referred to different departments over and over again; with incrementally larger waiting periods.
I know of a lot of people that I think are that shallow.
Few, nay... none of them are system administrators.
System administrator is a position of trust. The ones who aren't trustworthy either aren't system administrators, or they are very good at hiding their poor character.
Not that I necessarily think Snowden was a person of poor character; nonetheless, the NSA is charged with protecting secrets, and analytically speaking -- whether you agree with them morally or politically or not - folks like Snowden are a potential risk.
The fact that one admin did it casts some measure of doubt on the integrity of the lot of them.
Snowden is unusual.... but there must be more folks like that out there.
And he did sully the public image of System Administrators as a legitimate profession.
He didn't just hurt the country or himself, he hurt the trust of everyone in his former profession.
Now I don't necessarily agree he should go to jail; if he was a whistleblower -- calling out illegal NSA activity, then he should have the protection of the law. But the NSA has a responsibility to do what it reasonably can to prevent incidents like the Snowden leak.
Short of injecting employees with remote kill switches the NSA can flip at will to erase their brain or incapacitate them, until they can be captured --- if they have defected or been found stealing or leaking documents
... 100% of potential leakers are now 90% sure that they're going to lose their job anyway.
The catch is... the ones who are left, are going to be the ones that get to maintain the automated scripts; as a result,
the 10% who are left are probably going to wind up being the go-to folks, which ultimately suggests the administrative duties
and powers the 90% had will now be concentrated in the 10% --- so if one of those guys turns out to be bad, it may be even worse, and there will be fewer other human admins watching them plus possibly understanding dubios activity....
Meter each person's use of the photo copier -- via biometric and PIN number, put a barcode on each document, and have the copier log a record of each document/page that was copied.
Put similar scanning mechanism on paper shredders to log destruction of pages.
If someone over and over again is copying top secret docs, and it doesn't get shredded or logged into a secure area, then notify management
about the "outstanding" documents.
Oh, that's right, you don't have a good LDAP implementation, or group policies, or an update server, so how do you do it?
Group policies are a limited windows solution for solving a problem introduced by Windows. It's not in principal superior to the more general solution of using a tool such as cfengine or puppet to distribute configurations.
There are half a dozen good LDAP implementations; You can use Samba4; You can use OpenLDAP, 389 Directory server.
For ID mapping, NIS+; for Kerberos, MIT Kerberos, Samba4, or Heimdal.
RHDS; Redhat Enterprise IPA; FreeIPA.
FreeIPA is great.
Yes there is some policy functionality.
By the way.... the Kerberos and LDAP technologies were implemented on UNIX first and carried over to Windows later.
Windows never had a truly faithful implementation of them --- they always had to throw in some stupid dressed-up MS tweaks to bring in artificial incompatibilities.
There isn't anything really special about Linux that enables these tools to work, and I actually think the Windows Puppet agent gets off easy with NT services vs. init scripts with sketchy status commands, registry vs hundreds of different config syntaxes, and so on.
Really? Let me know when puppet allows me to login to a Windows server and type "yum install exchange-server":)
I have a couple of SAP, Exchange and other Windows servers I have to manage. They don't require any more babysitting that any of the linux boxes do. They're all VMs on Hyper V or Xen or ESX and I worry more about patching the host firmware than anything else.
I work with about 7 Exchange organizations; all deployed as VMs.
Hyper-V is nasty... careful not to get any of that stuff on you.
At any rate; these Windows services DO require more babysitting. Or at least, the admin team gets more "issue reports about them" than regarding any other Linux-based enterprise mail servers --- mostly though, the server works and the Outlook client side has issues, most issues deal with the Junk mail folder, Calendar, Calendar sharing, Public folders, or someone having a 20 gigabyte inbox and crashing Outlook against our strong (but ignored) advise to management to keep inbox quotas below 1gb, and require users to use the archiving system to access their 50gb photo collection, for stability, performance, and disaster recovery reasons: the client software, and the separate AD domain the client is occassionally using is something the Exchange admins have no control of, so all the need to "babysit" the server, is almost always to review (and reject), and lead back to the bright and narrow path --- some clueless beginner Windows server admin's request to allow 2 gigabyte attachments, start changing random Exchange settings, or apply some random registry hack they found on a forum somewhere, etc, etc.
In other words: I'm saying.... Windows servers do have more issues, but they are all the result of user abuse. Often user abuse that is sanctioned by clueless IT admin or management folks.
The problem is not the technology: it is the people who are allowed to decide how the technology is used, and admins and users alike relying on hearsay information and not following vendors' formal recommendations.
Slashdot Mirror
I don't have to prove that my imaginary die is perfectly balanced.
You would have to prove that your perfectly damaged die with independent trials is comparable to the OP's situation; which it's not.
The proposition you are making applies to only a specific kind of statistical argument. In the real world there are plenty situations where a previous outcome influences the next.
I will also add that in the next 50 days I am far more likely to get attacked by a tiger than in the past 50 days.... because I didn't get attacked by a tiger in the last 50 days.
That only works if you cheated and already observed the last 50 days.
If I asked you to go deep into tiger country, unprotected, disguised as the Tiger's prey, for 100 days, Or for 50 days.
Than with 100 days, your chances of getting attacked are larger.
Now; if you already observed 50 days, then you already know you weren't, so your chances of getting attacked during the next 50 are not necessarily any higher.
But they may be higher; since you will be wandering into yet further new territory, and there will be yet more time and more tracks left, for the tigers to find.
They may catch on to a trail you left during the last 50 days, so they may actually be more likely to find you on day 51, than on day 1.
The assumption that Tiger attacks are entirely random and not related to your previous days spent in places at risk, is not a reasonable assumption.
I would want to drive manually most of the time, but it would be great to be able to have a built-in designated driver for trips to the bar :)
As long as you have the physical capability of overriding the computer and driving it yourself -- then you have a possibility of having control of the vehicle; therefore, it is illegal DUI, even if the computer is driving it at the moment: the fact that you can override it at will makes it still an offense.
I suspect the law won't be revised to allow you to be drunk: until an inspection proves that a drunk driver cannot in a moment of weak judgement override their computerized vehicle.
Hey, I don't care if someone wants a self-driving car. As long as it doesn't hinder, re prohibit MY ability to buy and drive manual cars on the road.
I feel that's OKAY, but once self-driving cars are proven; the sale of new vehicles that can be manually operated should be restricted to specialty vehicles custom made or produced and sold in small quantities, and for commercial buyers that have special requirements.
And liability insurance companies should be allowed to provide discounts for the reduced risk to operators; with the costs shifted to operators of manual cars, additional licensing qualifications, and blackboxes required, to save lives, and allow society to benefit from the elimination of auto accidents that smart cars should result in.
it would be absurd for me to believe that I would likely get a one the next 100 time. No, for the next 100 rolls I should expect to get about 16 or 17 ones.
Dice rolls have an important statistical property that doesn't exist in this situation. Dice rolls are independent. One dice roll does not affect the outcome of another dice roll.
That's not automatically true: in general it is not true, and you have to prove that it is true.
Most dice won't be perfectly balanced.
It's possible that 250K miles is a magic number: you kept your car the whole time. Some mechanical defect exists in a large number of cars that cause them to break down on average in between 250K and 300K miles; often causing an accident.
In that case... after having driven 250K miles: your next 50K milles are potentially more dangerous than your last 50K miles.
That isn't how statistics work. The one accident in 250k is an aggregate of a lot of data. It doesn't mean that it is expected that after you get to 250k and beyond you're more likely to get into an accident.
If you have driven 300K miles, you are more likely to have gotten into at least one accident having driven that many miles than if you had driven 250K miles.
Arguably; your probability of an accident per mile driven is nonuniform and nonindependent; miles driven in a parking lot have a lower weight --- if your driveway is longer than other people's, then you might drive more miles with lower expectation of an accident.
When you are experienced and have driven less than 10000 miles in your life; I would argue, that you are more likely to get into an accident in those 10000 miles than the next 10000 miles.
I would also argue that miles driven in heavy traffic or bad weather are miles that you are more likely to have an accident in --- you could have driven 100K miles in ideal conditions, and gotten no accident, by driving less safely than the guy who drove 50K miles in horrid conditions, but got into an accident.
I do not recommend accident-free miles as a metric of driver quality.
I would argue, that a large number of miles driven suggests you are a better driver than the bottom 1/3rd, because of the road experience you got, having driven all those miles.
The more hazardous the conditions you drove more miles under with no accidents, and the more of those miles that were driven on congested high-speed roadways and bumper-to-bumper traffic, and other miles where accidents are most likely to occur -- the better a driver you are.
They could have just listed the fatality rates of the different countries; or provided a color-coded list.
For it to be useful as a map; it should be more granular, than merely painting every country the same color..... it should show fatality rates for states, provinces, counties, cities, and individual streets. Now that would make sense as a map.
Then I'd have to say you probably haven't worked much administering Windows networks of any size or complexity. That wasn't even true under Windows NT 3.51 when I worked with it in 1996.
What exactly are you doubting?
You have a domain controller blue screen.... happens all the time. Take a few minutes, maybe an hour for someone to get to physically reboot it; by the time it comes back up a JRNL_WRAP condition has occured, and SYSVOL replication is now officially borked, and group policy cannot be distributed.
Plus as luck would have it... there is frequently one form of corruption or another in the AD database or SYSVOL, so broken versions of the group policy files or AD database get spread around; because NTFS, has such poor resiliency and of the 100 sites in the domain each with their own domain controller.... one of them is bound to eventually have one form of data corruption or another....
Yep, and it works really, really well. Well enough that there are Windows machines sitting on the platforms of remote train terminals for years quietly doing their jobs without ever being cracked
If group policy is even used on those; then it's in spite of group policy -- not because of group policy.
Linux remote terminals would tend to have a self-contained configuration distributed from a central source, and redistributed to machines only when required.
Windows' reliance on domain controllers' continuous correct operation makes things fragile.
One of the top ten issues that Windows servers frequently develop is failure of replication, and in particular --- failure to apply or properly distribute group policy updates.
A rsync of some flatfiles is infinitely more resilient and reliable than what Windows does.
I might note that, if it weren't for the fact that the folks in Sales are addicted to their Outlook features, the entire rest of the company would be delighted to be rid of both Outlook and Exchange.
Seriously.... if that's the case; I would consider running a split-organization with Exchange for the Sales folks, and Postfix; Cyrus, or Zimbra for everyone else.
It is indeed possible to do that, even if all users have the same internet domain -- seeing that exchange has such a concept as an internal relay domain, where unknown recipients get pushed out to another server of your choosing.
def get_legislator_vote(Representative me, Bill b) :Yes
else
return :Present
end
end
if me.received_campaign_contributions_from_opponent?
if me.can_spin_to_constituents(b)
me.add_actions(:acquire_favors, b.supporters)
return :No
else
me.add_actions(:cash_in_favor_to_get_no_votes)
return :Dont_Vote
end
end
if b.bill_gives_government_more_power? && me.can_spin_to_constituents(b)
return :Yes
end
if b.bill_restricts_government_power?
return :No
end
if b.cnn_says_bill_addresses_current_crisis?
return :Yes
else if b.cnn_says_bill_will_cause_apocalypse?
return :No
end
if me.party == b.billsponsor.party :No :Yes :Amend :Present :Present :No :Yes :Present :No;
if me.received_campaign_contributions_from_supporter? if me.can_spin_to_constituents(b, b.supporters) return
if me.party == currentpresident.party
if currentpresident.opposes? b
return
else if currentpresident.supports?(b)
if me.canSpinToConstituents?(b)
return
else b.amendable?
else
end
else return
end
else
if currentpresident.supports
return
lse if currentpresident.opposes? return
else return
end
else
return
end
end
If you have got 50,000 messages in your Inbox, you will probably experience some bad performance, even with Gmail
Blah.... bad IMAP performance, as in your mail client will begin to have trouble. Of course... the Gmail web interface is marvelous, and you can probably have half a million messages in your inbox, as long as you don't want an IMAP software program to index that and download the headers of all those messages.
About half of the folks in my group use Thunderbird with the Exchange server, and the crashes and other problems hit randomly - sometimes it's 1/2 Outlook and 1/2 Thunderbird users whose mail is not working because the remote server is having a bad hair day.
I can understand Thunderbird users having issues. Exchange's implementations of POP and IMAP are poor and should be avoided, plus the default throttling policies may not be adequate to stop some runaway IMAP clients causing serverwide issues. I can't accept "remote server is having a bad hair day"; either it was misconfigured, sized with too few resources (Sufficient IOPS at low-latency, RAM, Memory, CPU, Disk), being used in a way above or outside the scale that Exchange is designed for, not being adequately protected against abusers or misbehaving clients, or there is a hardware issue, or problem with a service it's dependant on such as AD. Bugs in Exchange are possible as well, but if experienced, the ones that really matter show up in the logs as more than "having a bad hair day".
I'll accept that a 20GB remote inbox might have indexing & searching lags, if there isn't a fulltext index at the other end, but that's a different issue.
The PST/OST file format on the client is prone to corruption at those sizes, which means a big amount of downtime for the end user to redownload 20GB of data.
The bigger a mailbox gets, the more storage I/O capacity is required simply for accessing that mailbox. Which means, that the entire server will run out of precious expensive IO resources with increasing speed, the bigger your users' mailboxes get. Although.... to some limited extent, you can counter this by having extra RAM on your exchange server, above and beyond Microsoft's standard recommendations of ~50MB per mailbox, for extra Gigabytes of active mailbox database storage you use, by multiplying that by a factor, and requiring 200gb of RAM for the Exchange database server, instead of 32gb.
It also means that you can essentially have 10 users per database until you reach Microsoft's recommended database size maximum of 200GB; in the event that you have a much larger database, the delay required to repair or restore from backup can easily extend downtime to an unacceptable duration.
If your users want to keep their old mail, use an archiving system that was designed with an architecture suitable for operating at that scale.
New-MailboxRepairRequests and various other operations for large mailboxes result in unacceptable downtime for the user, when these are supposed to be routine, mostly transparent activities. Bringing a system to conditions where that system cannot be made to recover gracefully from occassional errors or problems is a bad situation to have.
I'll also note that my GMail account probably has 30,000 emails in it right now.
It's not fair at all to compare Gmail to Exchange. The architecture is completely different. Gmail's infrastructure is a much more scaleable system, they have loads of cheap storage in a proprietary magic Google clustering system that few can get close to matching ---- however, it's also managed by someone else, AND it's distant -- so if you need your mail fast, Gmail's not necessarily a good idea.
On the other hand, if you need large attachments, Gmail's probably a great idea.
Exchange's selling point is not that it's a scaleable system. It's that people are used to Outlook and like its features.
If you have got 50,000 messages in your Inbox, you will probably experience some bad performance, even with Gmail
And the note about inboxes - in our business we deal with a LOT of email that is several megabytes per message.
This is likely to weigh heavily on Exchange, best practices are to reject them -- to have a reliable system; remember E-mail is a store and forward system.
Messages that are more than a meg or so are a potentia
All? I guess there's no point in installing all those updates then.
Sure... install EMET, enable all the mitigations, rubber stamp the latest Windows 7 release; make sure your users don't install Java or Flash (ban the plugins), and you're pretty much as good to go, as if you installed every update --- your users will still manage to get malware on there, and not updating doesn't make it more likely or less likely.
For bonus points, add an AV or application whitelisting solution ala Bit9. The antivirus has about as much chance catching a malicious file as a coin flip though. Just a fact of life -- AVs used to be better, but have become THAT ineffective.
That's easy enough to say for someone who hasn't had to deal with 10 billion years of other lifeforms and their bullshit.
I don't care about that shit. I'd just ignore it all and play video games for a few eternities, after cashing in the dollars in interest and stock dividends earned over my first 200 years of immortality.
What I wouldn't want to do is live forever but still age, live forever as an old person, or have the risk of being injured in pain, hungry, trapped, or disabled and still live forever -- but in some horrible predicament; or be a person that has to put up with the ailments of old age or sickness.
I would rather be an immortal 6 or 7 year old girl who is uninjurable, invincible and can never feel pain or discomfort, than live forever as a 60, 70, or 80 year old, who society allows to do about whatever they want.
In other words: cryonics on old people is insane; unless we can stop aging. It makes more sense for the young instead, heh....
I sure don't: After a couple of centuries, I'd get bored, and I don't really feel like going around insulting the universe
I do want to exist and be sentient and mindful forever, and religion promises me that I will.
Do we really need technology to achieve what God has already promised us?
If that happens, I wonder if they'll be able to keep their employees around, or if all the good ones will just leave.
If they'll bail out just because of some price swings of publicly traded stock, or they were just there because of insanely high valuations, then they are not very good employees.
You don't work for a company based on its stock price. You invest in a company's shares based on its stock price being below what you believe the shares will be worth; or what you believe those shares will eventually be worth relative to the market and relative to your opportunity costs in regards to other options for investing your cash.
Yes, that problem of allowing everyone in a company to have a computer on their desktop.
No.... the problem is the Windows registry; which without group policy makes distribution of standard settings difficult.
Group policy is nothing more than a system for deploying canned registry hacks to specified computers, based on which checkboxes someoned ticked and saved to a GPO file.
P.S. Free speech or making an offensive movie are parole violations. Or at least..... well, other things that would otherwise be overlooked become parole violations, after the exercise of said free speech.
You don't like someone's political views? Get them arrested on same fake charge; get them parole.
But if they dare speak out again against our overlords, you can be sure it will be a parole violation
Microsoft will always be there to help.
Yeah... after spending 24 hours on hold, then waiting another 24 hours for a callback, and getting referred to different departments over and over again; with incrementally larger waiting periods.
Do you really know that many people that shallow?
I know of a lot of people that I think are that shallow.
Few, nay... none of them are system administrators.
System administrator is a position of trust. The ones who aren't trustworthy either aren't system administrators, or they are very good at hiding their poor character.
Not that I necessarily think Snowden was a person of poor character; nonetheless, the NSA is charged with protecting secrets, and analytically speaking -- whether you agree with them morally or politically or not - folks like Snowden are a potential risk.
The fact that one admin did it casts some measure of doubt on the integrity of the lot of them. Snowden is unusual.... but there must be more folks like that out there.
And he did sully the public image of System Administrators as a legitimate profession. He didn't just hurt the country or himself, he hurt the trust of everyone in his former profession.
Now I don't necessarily agree he should go to jail; if he was a whistleblower -- calling out illegal NSA activity, then he should have the protection of the law. But the NSA has a responsibility to do what it reasonably can to prevent incidents like the Snowden leak.
Short of injecting employees with remote kill switches the NSA can flip at will to erase their brain or incapacitate them, until they can be captured --- if they have defected or been found stealing or leaking documents
The catch is... the ones who are left, are going to be the ones that get to maintain the automated scripts; as a result, the 10% who are left are probably going to wind up being the go-to folks, which ultimately suggests the administrative duties and powers the 90% had will now be concentrated in the 10% --- so if one of those guys turns out to be bad, it may be even worse, and there will be fewer other human admins watching them plus possibly understanding dubios activity....
Meter each person's use of the photo copier -- via biometric and PIN number, put a barcode on each document, and have the copier log a record of each document/page that was copied.
Put similar scanning mechanism on paper shredders to log destruction of pages.
If someone over and over again is copying top secret docs, and it doesn't get shredded or logged into a secure area, then notify management about the "outstanding" documents.
Oh, that's right, you don't have a good LDAP implementation, or group policies, or an update server, so how do you do it?
Group policies are a limited windows solution for solving a problem introduced by Windows. It's not in principal superior to the more general solution of using a tool such as cfengine or puppet to distribute configurations.
There are half a dozen good LDAP implementations; You can use Samba4; You can use OpenLDAP, 389 Directory server. For ID mapping, NIS+; for Kerberos, MIT Kerberos, Samba4, or Heimdal. RHDS; Redhat Enterprise IPA; FreeIPA.
FreeIPA is great. Yes there is some policy functionality.
By the way.... the Kerberos and LDAP technologies were implemented on UNIX first and carried over to Windows later. Windows never had a truly faithful implementation of them --- they always had to throw in some stupid dressed-up MS tweaks to bring in artificial incompatibilities.
There isn't anything really special about Linux that enables these tools to work, and I actually think the Windows Puppet agent gets off easy with NT services vs. init scripts with sketchy status commands, registry vs hundreds of different config syntaxes, and so on.
Really? Let me know when puppet allows me to login to a Windows server and type "yum install exchange-server" :)
I have a couple of SAP, Exchange and other Windows servers I have to manage. They don't require any more babysitting that any of the linux boxes do. They're all VMs on Hyper V or Xen or ESX and I worry more about patching the host firmware than anything else.
I work with about 7 Exchange organizations; all deployed as VMs.
Hyper-V is nasty... careful not to get any of that stuff on you.
At any rate; these Windows services DO require more babysitting. Or at least, the admin team gets more "issue reports about them" than regarding any other Linux-based enterprise mail servers --- mostly though, the server works and the Outlook client side has issues, most issues deal with the Junk mail folder, Calendar, Calendar sharing, Public folders, or someone having a 20 gigabyte inbox and crashing Outlook against our strong (but ignored) advise to management to keep inbox quotas below 1gb, and require users to use the archiving system to access their 50gb photo collection, for stability, performance, and disaster recovery reasons: the client software, and the separate AD domain the client is occassionally using is something the Exchange admins have no control of, so all the need to "babysit" the server, is almost always to review (and reject), and lead back to the bright and narrow path --- some clueless beginner Windows server admin's request to allow 2 gigabyte attachments, start changing random Exchange settings, or apply some random registry hack they found on a forum somewhere, etc, etc.
In other words: I'm saying.... Windows servers do have more issues, but they are all the result of user abuse. Often user abuse that is sanctioned by clueless IT admin or management folks. The problem is not the technology: it is the people who are allowed to decide how the technology is used, and admins and users alike relying on hearsay information and not following vendors' formal recommendations.