« first day (1459 days earlier)      last day (3487 days later) » 
00:00 - 20:0020:00 - 00:00

slm
12:21 AM
@jimmij - why did you delete you A to the meta Q?
Opposing opinions are fine on meta Q's. They're actually important b/c they show that things were discussed when the community is trying to reach a consensus on important issues.
@mikeserv - same for your A to the meta Q, why the delete?
 
12:47 AM
@slm Everything has been already said, I see that decision has been made and it is irreversible (although nobody wants to say so). Anyway since mikeserv deleted his posts I don't see a reason to fight alone for this anyhow lost case.
 
@jimmij Nothing at all has been decided! That's the whole point of the meta post, to promote discussion. If dissenting points of view are deleted, only one point of view will be heard.
 
slm
1:37 AM
@jimmij If you truly have concerns that's the time to bring them up. If other's agree w/ you then they'll chime in on your A or on mike's. Since you're new to the SE sites you might want to take a look at the various discussions on meta.stackexchange.com or meta.stackoverflow.com. There can be some heated discussions but they're important so that everyone understands other's positions and they help the sites to function.
 
 
12 hours later…
1:30 PM
VTC, people.
0
Q: What is the difference between bash man, info and help commands?

REACHUSHow are the: man, info and help commands different? Which one should I use in which context?

 
slm
@FaheemMitha - closed
 
@slm Thanks.
Hi. I was just sent a link from Debian to freexian.com/services/debian-lts.html
I notice that the hourly pay mentioned there is 85 euro per hr, which i see works out to around $107 per hour. This strikes me as quite nice pay. Is this typical for this kind of work in Europe?
 
slm
2:09 PM
That seems very nice (40*52)*100 = 208k. But everything is based on supply/demand. A good C/C++ developer is always in high demand. One that knows kernel code even more so.
I would say that's a going rate, maybe even a little low, I know of many contractors that do C/C++ that get upwards of $150/$200 per hr.
 
2:42 PM
I am confused. Is https not an application layer protocol?
Wow, it seems it a mixture of application and session layer protocol as discussed here.
 
@Gnouc, dvted.
 
slm
3:37 PM
@Gnouc - DV
 
3:51 PM
@Gnouc you want us to downvote your answer?
 
@Braiam: What do you mean?
 
the link you gave, points to your answer
 
@Patrick - I am 100% sure what you say is wrong. I am running my own tftp server and I know how it works. The tftp protocol must be extremely basic/trivial, so that clients can be simple. But the server does not need to be basic. Indeed my tftp server decides what boot image client gets, depending on IP. I am using tftpd-hpa. — Martin Vegter 1 min ago
... and that's why you're here asking for our help, because you know what you're doing...
people that come on here asking for help, and discard the information they're given irritate the crap out of me
 
...
at those time I just drop a link with the documentation...
you think this is fine? networkboot.org/fundamentals
 
4:01 PM
unfortunately the documentation is RFC. But i have a feeling he's not going to understand RFC, and I can't explain it in a comment (and I don't think it's relevant as an answer)
@Braiam Doesn't really cover what he's claiming
He's claiming that he makes a single tftp request, and the server sends him a bunch of files
and that the server ignores the requested filename and sends what it wants to instead
 
darn... the wiki's only explain "how to" but not "how is done"
 
what is this option supposed to do?
options iwlwifi 11n_disable=1 wd_disable=1
 
slm
@Patrick let them whither, don't waste your time!
@Patrick - that user can be very stubborn at times, I generally shy away from A'ering him b/c simple A's invariably turn into me regretting that I offered them up.
 
4:20 PM
@Ramesh the first one should have never existed
 
But the first one was the actual solution.
 
@Ramesh ... not only because its link-only, but seldom should you post answers you're not confident are correct.
 
@derobert Oh, if only that were reality on this site :-)
 
hmm, but that answers the question and I was sure that was correct.
 
@Patrick Yeah. But as long as we're talking about what should be done :-P
 
4:22 PM
@Ramesh do I have to click the link to solve the problem?
 
I could have rather copy-pasted the info from link.
But I did the testing in my machine and added the details. What's wrong with that?
 
do the original answer solve the problem without clicking the link?
 
And moreover I have cited the site. It's not like I am claiming that it is my solution.
So is it ok if I copy paste the contents from the link without actually testing?
 
> Yes. You can use the file at /etc/pam.d/sshd. For more details, you could see this link. I will test it and update in a while.
^ do that solves the problem?
 
@Ramesh it sounds like (from that) you're not actually sure it'll work. But that for some reason you posted an answer anyway...
That's in addition to the link-only problem.
 
4:27 PM
@derobert, I knew the solution would work. But the problem was I wanted to test as well which would prove that the information from the other source is correct.
So what's wrong with that? I mean it is not like I am not going to add information.
 
@Ramesh why aren't you answering my question?
 
So if I just copy pasted the link info, does that work?
I didn't do in this case, but will that work for future questions?
 
@Ramesh does your original answer solves or not the question posted by the OP? just answer that
 
I did not copy paste the link info. If I have copy pasted the link info, yeah it does.
 
@Ramesh "I will test it and update in a while" sounds like you weren't confident it'd work.
 
4:30 PM
@Ramesh there's the problem. It should have solved the question since the start
 
@derobert, I accept that. I just wanted to test it in my system. Is it wrong?
 
@Ramesh testing has nothing to do with what I'm trying to tell you
 
Testing your answer is good. Preferably before posting it, though.
 
So, I believe I should have put in the current version of my answer as the first answer, is that it?
 
Yes
 
slm
4:32 PM
@Patrick - I suspect Martin's getting confused by the capability to selectively install different images on different servers from tftp like something seen in Cobbler. I've only ever dealt with Cobbler's use of TFTP so I'm not sure how other implementations may make use of it.
0
A: copy files from tftp server

slmI'm most familiar with how Cobbler sets up TFTP so I'm not sure if this is relevant but I'll offer this info up anyway. Cobbler sets up a bootloader using pxelinux.0 as @Patrick explains in the comments as well as @msw. But it also sets up a corresponding pxelinux.cfg/ directory along side it tha...

 
hey @derobert. long time no see.
figuratively speaking, of course.
 
Yep, no more conventions until Feb.
 
No ComicCons, then?
 
Man, it's so frustrating when someone asks whether they can modify my resume :(
 
slm
@Ramesh - ???
 
4:36 PM
I second that notion. ???
 
@slm It's just some consulting firm suggesting to make changes in my resume.
They basically want to fake experience :(
 
slm
If ppl have votes to spare, mikel is close to 20k
 
@spider, well, that was not correct and I accept it was wrong of me to post an answer without actually testing. I had discussed in the chat regarding this and I was told it is not the right way to do things. I am sorry for that :) — Ramesh 1 min ago
 
slm
@Ramesh - this article is long but funny and sad about the entire interview process for technical vs. managerial jobs works.
 
@slm sadly most of the companies I interviewed with look at external factors such as sponsorship, years of experience, certifications, expertise in some corporate level tools etc without actually evaluating whether the candidate is a good fit. :)
 
5:24 PM
@Braiam @Ramesh @slm you should indeed downvote @Gnouc's answer, it's wrong. Archemar's answer is mostly correct.
 
@Gilles: Can you prove?
 
@Gnouc if it worked the way you described, the first iteration would not be printed
 
while not? The first time, a[$0] is not define, ++ increment it and return a null value
 
there must be a thousand threads on SO about this
 
null negate --> true?
 
5:28 PM
precedence has nothing to do with order of evaluation
The value of a[$0]++ is the old value of a[$0]. The incrementation hasn't happened yet.
The ! operator is applied to the old value. !a[$0]++ has the same value as !a[$0].
 
slm
I've edited both the A's so you can undo any voting that was done in error.
 
Really? The doc said Increment lvalue, returning the old value of lvalue as the value of the expression?
 
The difference is that !a[$0]++ also increments a[$0]. It's equivalent to {tmp = a[$0]; a[$0]=a[$0]+1; !tmp} or {tmp = !a[$0]; a[$0]=a[$0]+1; tmp} (and you aren't allowed to observe the difference between these two)
@Gnouc returning the old value
 
yes, as your explain, the first iteration will not be printed
simply, ++ has higher precedence than !
a[$0]++ will be evaluated first
 
@Gnouc this has nothing to do with precedence
 
slm
5:34 PM
@Ramesh @Braiam ^^^^^
 
21
A: Why does "++x || ++y && ++z" calculate "++x" first, even though operator "&&" has higher precedence than "||"

JayUnwind, R and others have explained what really happens. So let me just add: The premise of your question is faulty. The fact that the && has higher precedence doesn't mean that operands that surround it must be evaluated before any operands in the expression with lower precedence. Even where th...

 
@Gilles: So do you think "awk 'BEGIN{a="";print !++a}'" will print what?
 
@Gnouc 0
 
So why? Is not it because ++ has higher precedence than !?
 
THIS HAS NOTHING TO DO WITH PRECEDENCE
 
5:37 PM
prove it please?
 
@Gnouc look at the parse tree (which I am NOT going to draw for you)
then at the evaluation rules
 
@slm That's nice money. Of course, the drawback is that you have to write C/C++. Are these people all located in the US?
 
slm
yes
 
I see the gawk source code
 
@slm Why kernel code?
 
5:39 PM
It said: ++ increment by 1 [expression (++a) gets the current value of
a and adds 1 to it, stores that back in a, and returns the
new value; expression (a++) gets the current value of a,
adds 1 to it, stores that back in a, but returns the
original value of a]
If you said ! is applied first, then the first iteration print nothing
since when the last value of expression return is a false value
 
@Gnouc yup: return the original value
 
It's a fail value ok?
 
That's why !(a[$0]++) returns the negation of the old value of a[$0]
 
As you said, ! is applied first ok?
so !a[$0] is evaluated first, it return true value
after that, ++ increment a[$0]
and return the old value of a[$0]
which is a false value ok?
 
@Gnouc More precisely, it's applied to the old value. AFAIR in awk you can't observe in which order the ! operator and the assignment to a[$0] happen (they could even happen in parallel)
 
5:42 PM
So the final result of whole expression is a false value?
 
Yes
er wait, which expression are we talking about
 
So why the first iteration is printed, since when it return a false value?
!a[$0]++
 
you changed expressions mid-stream
@Gnouc this one is true the first time, false subsequently
 
So which is the final result of expression?
 
@slm i actually think this is a reasonable question
0
Q: Why didn't Gnu Info succeed man?

shirishAs per my knowledge/understanding both help and man came at the same time or have very little time difference between them. Then GNU Info came in and from what I have seen is much more verbose, much more detailed and arguably much better than what man is. Many entries even today in man are crypti...

i also sort of know the writer. he's active in Debian (in India).
 
slm
5:44 PM
Sounds primarily opinion based to me.
Can others please comment on whether this Q should be reopened or left closed as opinioned?
 
@slm There is quite a lot one can say about why into did not take off.
I could make some comments. I'm not sure if I feel competent to attempt a full answer.
 
slm
@FaheemMitha - right but it's all conjecture and opinions which makes it a bad Q for the site
 
@slm Hmm. Not sure about that. There might be some actual reasons why it didn't work.
 
slm
I only blocked it quickly b/c I didn't want a bunch of A's getting filled out. How would he accept one too?
 
@slm A historical overview. Do you really think this will get a lot of answers?
 
slm
5:47 PM
See what others think, if others want it opened they can override me or I or another mod can reopen it
 
@slm Fair enough.
 
slm
have you not seen what others have written here about is linux unix etc etc.
 
@slm Um, what?
 
slm
90
Q: Is Linux a Unix?

Shawn J. GoffSo, there are lots of different versions of Unix out there: HP-UX, AIX, BSD, etc. Linux is considered a Unix clone rather than an implementation of Unix. Are all the "real" Unices actual descendants of the original? If not, what separates Linux from Unix?

these types of Q's though this is a good one, but I'm showing it b/c it elicited a lot of A'ers.
 
@Gilles: I'm waiting your answer?
 
5:49 PM
@slm I see. But that's more of a general interest question.
 
slm
I highlighted this to show that people have a lot to saw on open ended Q's when given the chance.
 
@slm Yes, I see.
 
Hi
 
slm
pass that one around, it easy enough to reopen if it's warranted.
 
I have a script hanging at boot time, how can I cancel it, it runs before I even login
I tried ctrl + C /D
 
slm
5:51 PM
you could boot into single user mode and disable it from running
 
@slm
how do u do that
 
slm
5
A: Booting from command line

slmTake a look at this tutorial titled: Learn Linux, 101: Runlevels, shutdown, and reboot. You can change from runlevel 3 (non-GUI Desktop) to GUI Desktop (typically runlevel 5) with the following command: $ telinit 5 You can see what runlevel you're currently in with the runlevel command: $ run...

assuming it's using SysV init scripts, what's the distro?
 
its ubuntu
 
@slm If you mean the info thing, I doubt anyone cares enough. I'm actually somewhat interested in the issue. It's a mildly interesting historical issue.
 
8.0.4 debian
 
5:54 PM
@meda which of those?
 
slm
yeah which one?
 
64 bit
what do you mean wchich one
@FaheemMitha
 
@meda you said ubuntu. then you said 8.0.4 debian. Which isn't a release number, btw.
 
slm
debian or ubuntu?
 
Right. it has to be one of the other, can't be both
 
5:58 PM
I thought Ubunutu is from debian distro
 
but then its ubuntu if anything
 
slm
@Gnouc I'm a awk novice, was letting you 2 work it out. The only other person I've seen that is proficient with awk is hauke, or glenn jackman
 
@meda lsb_release -a
 
slm
@FaheemMitha - can you assist meda?
I need to get back to work
 
6:01 PM
@slm I don't really want to.:-) But I'll take a look.
 
@Gilles: Are you there?
 
@meda did you have a question?
 
slm
@Gnouc - be patient he's on like 20 other sites
 
@slm: He does not make the explanation clearly, and give a down vote.
 
slm
@Gnouc - it seemed as though he was trying to explain it above in chat.
I'm not sure where the disconnect is, only that you two were discussing it.
He's usually pretty gracious about explaining things so that ppl understand it.
 
6:15 PM
@slm: I don't think so. He said I'm wrong, but can not prove that
Now he left :)
 
slm
@Gnouc - it would seem that way...
 
/me imagines Gilles with 8 arms like Dr. Octopus.
 
Stéphane Chazelas always give me the reason, the explanation when he give me a down vote
 
slm
@Gnouc - if you can put together a small example that shows why you're right you can post it as a Q and then A it and then reference it in your other A to solidify your position
@Gnouc - Looking through I saw that Gilles had done the same, but above it looks like he might have been confused with the expressions in the discussion?
Best way to prove you're right is to prove the other position is wrong.
 
Expressions in awk

Expressions describe computations used in patterns and actions. In the following table, valid expression operations are given in groups from highest precedence first to lowest precedence last, with equal-precedence operators grouped between horizontal lines. In expression evaluation, where the grammar is formally ambiguous, higher precedence operators shall be evaluated before lower precedence operators. In this table expr, expr1, expr2, and expr3 represent any expression, while lvalue represents any entity that can be assigned to (that is, on the left side of an assignm
 
slm
6:19 PM
Can you construct some small examples that show what happens with that code?
 
@slm: He said "THIS HAS NOTHING TO DO WITH PRECEDENCE"
If you read the doc, you can see he was wrong
 
slm
yeah I read that. It's not in my wheel house unfortunately. I'd go and construct examples to show that what I think is happening is.
What about this url he posted: stackoverflow.com/questions/3700352/…
The conversation about it starts here
1 hour ago, by Gilles
@Braiam @Ramesh @slm you should indeed downvote @Gnouc's answer, it's wrong. Archemar's answer is mostly correct.
You can pin it on the side so others that may know more about awk can chime in as well.
 
@slm: I don't think it's relative, since when that question about C, not awk
That question is also in different case
 
slm
@Gnouc - yeah I'd have to go and do a fair amount of research to make sure I understood the situation before getting involved
My mind is in Perl and Ruby right now so C and awk are in the backseat
@Gnouc - reference this thread in your A and dis-spell that you're A is wrong.
 
@slm: What do you mean?
And if you talk about Perl, you can see:

perl -e '$c=1;print !$c++'
print nothing :)
 
slm
6:33 PM
@Gnouc - Gilles left comments basically saying your A is incorrect, so highlight that you attempted to discuss it and it would seem he was incorrect based on the conversation in chat. Add this as an update to your A at the bottom so that others that read it can make their own conclusions or shed more light on it.
@Gnouc I understood what you were saying before, but not the implications wrt awk. 8-) Thanks for dumbing it down into Perl though. 8--)
 
My answer still getting down vote, but I don't care much about that
 
slm
I had reversed my votes based on Gilles feedback.
I didn't want to keep bouncing around since I've now flipped my voting when you 1st suggested it and now back
was waiting for you two to work it out
 
I will delete my answer, now I must go sleep
bye all
 
@slm Interesting problem, I'm not expert so I didn't vote, but I think this question should be closed as a duplicate of next "canonical" question (yet to ask) about awk operators, precedence, and increments.
 
slm
@Gnouc - pls do not delete it
@jimmij - feel free to write it
canonical Q&A can be written by anyone
 
6:52 PM
It was only a joke, I like questions like that where one can discover problems on specific example, without reading "canonical" manuals (which anyhow nobody remember by heart)
 
slm
I meant nothing other than to write any canonical that you see fit.. They are not stones handed to us on a mountain, they usually represent a group within the community that is fed up with the same Q's being asked and wants to pull together to do something more than just machine gun answer the same old same old over and over.
Q's that push the envelop of everyones understanding are good too, and hence why many hang around here, to learn the things that they don't know, rather than recycle the things that we already do know.
or are easily accessible via a google
 
Is it correct to say bridges operate in layer 1 and layer 2 where as routers operate in layer 3 of the OSI model?
 
slm
Switches /Bridges/Wireless Access Point are found in the Data Link Layer
#2
 
So it is layer 2 alone.
Thanks @slm.
 
slm
yeah, since 1 is the actual wires
 
7:07 PM
so firewall rules come place for layer 3?
In fact it can go from layer 3 to layer 7 as discussed here.
 
slm
Seems like it's def. 3 and 4. It can operate higher too in looking at the diagram in the link I sent you. You might want to ask this on the networking SE site
we have one of those don't we?
 
So for a Linux sysadmin job, is it necessary to dig so much details about networking?
 
slm
I don't think so
 
I feel it is interesting which is why am looking at them. :)
 
slm
it's good to know but the reference to the OSI only seems to ever come up here.
 
7:13 PM
yeah, the link for OSI layer that you had given is really good.
 
slm
it's good to understand how to read the osi model and how it pertains to things but it's a reference that you go and consult when needed
 
@Ramesh depends on what you want to do for your job. basic entry level, no. More advanced level, absolutely
 
I am just a beginner with 1 year experience (in fact just 1 year experience with Linux) :)
 
slm
@Ramesh when you asked about a bridge did you mean a network card in bridge mode?
or an actual bridge networking device?
 
@slm, I am not sure about the difference.
I am just learning things.
 
slm
7:15 PM
Network bridging is the action taken by network equipment to create an aggregate network from either two or more communication networks, or two or more network segments. Bridging is distinct from routing which allows the networks to communicate independently as separate networks. A network bridge is a network device that connects multiple network segments. In the OSI model bridging acts in the first two layers, below the network layer. There are four types of network-bridging technologies: simple bridging; multiport bridging; learning, or transparent bridging; and source route bridging. == See... ==
A network bridge is a network device that connects multiple network segments. In the OSI model bridging acts in the first two layers, below the network layer.[3]
The card I'm not so sure about
I would guess it's 3
 
I was looking at this command which is when I felt confused.
arp -i wlan0 -Ds 192.168.1.9 wlan0 pub
Is it something getting executed at layer 2?
 
slm
arp is the table that is maintained w/ all the mac addresses in it
shows the map of mac -> ip
doing a arp -a shows the whole thing, the arp -i wlan0 shows only ones on that i/f
the rest I'd have to double check man pages but is likely showing only a further subset from a specific network or subnet
that example is right from the man page
   This will answer ARP requests for 10.0.0.2 on eth0 with the MAC address for eth1.
-Ds looks to do this: arp -Ds `address` `ifname` form. In that case the hardware
address is taken from the interface with the specified name.
 
Ok, so we are trying to create a bridged networking in layer 3? sorry if am asking something stupid :)
 
bridging only occurs in layer 2, routing occurs in layer 3
 
slm
so the bridging of 2 NICs is layer 2?
 
7:24 PM
correct
 
So, arp protocol is different from arp command?
 
it is a requisite of a subnet to be in a single broadcast domain
 
slm
There's also tuns and taps which I've yet to work out how they work, but you seem them with virtualziation now
 
The reason am asking is because in this link arp seems to be present in layer3.
 
slm
@Patrick right that part I knew
 
7:26 PM
@Ramesh arp is what maps layer 3 onto layer 2
 
slm
you'll often see the calculators online for helping to workout a given subnets particulars such as the broadcast addr
 
@slm I found this tutorial extremely helpful in calculating those stuffs.
 
slm
classes are no longer used as I understand it, they were replaced by cidr
 
@slm oh ok. I will get acquainted to that now :)
 
slm
Classless Inter-Domain Routing (CIDR, pronunciation: /ˈsaɪ.dr/ or /ˈsi.dr/) is a method for allocating IP addresses and routing Internet Protocol packets. The Internet Engineering Task Force introduced CIDR in 1993 to replace the previous addressing architecture of classful network design in the Internet. Its goal was to slow the growth of routing tables on routers across the Internet, and to help slow the rapid exhaustion of IPv4 addresses. IP addresses are described as consisting of two groups of bits in the address: the most significant bits are the network address, which identifies a whole...
I hadn't realized that that had changed until last year 8-)
someone mentioned it in one my A'ers heree
 
7:34 PM
Is it possible to know where a packet is getting dropped?
I mean, can we use tcpdump command to find it?
 
slm
dropped how?
 
you can use tcpdump to determine if the packet is being sent. but if it's not, it won't tell you why
...or who dropped it (and there are numerous possible culprits)
 
sender A sends a TCP packet to sender B and in between the packet gets lost somewhere. So can we determine where exactly the packet gets lost in the middle?
 
slm
yeah determining the source of where it was dropped is very trick to pin point
it could be any number of things at various levels doing the dropping
wire, NIC, firewall
switch in b/w
 
@Ramesh define "sender A". are we talking at the application level, the kernel level, or the box level?
 
7:38 PM
@Patrick It's application level.
 
@Gnouc I am now
 
slm
@Gilles - he's gone to bed now
 
@slm also, I (well, the man behind the curtain) also am outside SE
 
slm
did he delete his A
 
@Ramesh there's no one way to determine who dropped it then. And sometimes you can't even add inspection points along the way to see "did it make it to this layer".
 
slm
7:40 PM
@Gilles I know, I figured something else came up, I told him to be patient but he was at the end of his day
he's taken your absence as a sign that you were wrong and were ducking him now
 
@Patrick Thanks for the info.
 
I don't think I have anything to add to that conversation anyway.
 
slm
Yeah I figured you just disengaged at that point, it wasn't going anywhere you already said your piece
 
We are failing at teaching programming. (I guess I should say we since I was a TA for a few years.) Syntax is a detail. Getting hung on syntax issues like precedence is failing to see the forest for a twig.
 
@Ramesh However, in a large number of cases, you can add iptables rules to log the packet. For example, put a rule in the prerouting chain, then the postrouting, etc.
 
slm
7:44 PM
I suggested to him to try and construct an example that would highlight why he thought he was right
lol
 
@slm he tried. But the example he came up with was different, and I didn't manage to explain to him why it was different.
 
So in cloud environment, all of these things comes pre configured?
I mean, the iptables rules, the ports that needs to be available etc?
 
@Ramesh what things?
not generally no
 
Just the OS image then?
 
more or less
 
slm
7:47 PM
Wasn't sure who else could help him, I suggested Glenn.
 
site specific question. Why do we have distribution-choice tag when most of those questions would be either opinion based or too broad?
 
Oh hey, I have a riddle for networking boffins
I don't have access to the machine in question, so I can't easily provide more information, but I can try and forward
 
For instance, this is well written but still I feel it could be opinion based.
 
@Ramesh I wouldn't mind declaring them off-topic (specifically, as opposed to other kinds of software recs), or at least put a heavy burden on them to prove their worth. I use as “I don't want to see this question”, it's my one ignored tag.
anyway, my networking issue
This is on a perfectly normal home PC behind a DSL router. Ubuntu 14.04, freshly updated, but I'm told the same problem was seen on 12.04
The user has a website with a hosting provider. She uploads files over FTP. She also has an SSH account (not root) on the same machine as the FTP server.
Suddenly, FTP uploads stopped working. It used to work, then it stopped. The FTP client would create a 0-byte file and hang.
Everything else works normally, including FTP downloads and web browsing.
I investigated and found that files up to 1452 bytes uploaded correctly.
Ok, that's easy. MTU problem.
Question is, what could have changed so that it worked for years and suddenly stopped working?
 
ICMP cant fragment messages getting blocked is my first guess
 
7:55 PM
Reducing the MTU from 1500 to 1492 did solve them problem.
I checked, and the machine did receive ICMP fragmentation needed
I don't understand why it didn't resend the packets
 
tcpdump showed them, but did they make it through iptables?
(the icmp packets)
 
@Patrick wide open
 
:-/
 
but the real mystery is:
I logged in over SSH, to see if SSH also exhibited MTU issues. Same server IP address.
The mere act of logging in over SSH made FTP work for a few minutes.
????
 
slm
did different ftp clients exhibit this issue?
 
7:58 PM
yeah, you're off in the land of "i'd need to see packet captures and a shitton of data"
 
MTU=1500, FTP uploads over 1452 bytes don't work and result in ICMP fragmentation needed. Open an SSH connection, same MTU setting according to ifconfig, and suddenly the FTP upload gets through.
@slm I tried both Filezilla and lftp, same behavior
 
slm
I'd also be suspicious of her router/switch that's doing the NAT
is it older, what model?
 
I don't know the model. Popular home router/DSL.
 
if the box received an ICMP can't fragment message, and didn't reply back with a smaller packet, that just screams filtering. That's at the kernel level, the app isn't even aware that's going on.
 
00:00 - 20:0020:00 - 00:00

« first day (1459 days earlier)      last day (3487 days later) »