Nyhetsflöde

Is there any examples of previous projects done by students?

Where should we submit the project proposal?

"2014 September 22 (after the lecture)" What time exactly? 10:00 or 23:59, or something else?

The number of points on the previous projects would also be helpful.

Examples are: chat program, online game (based on earlier course project on stand-alone game), non-concurrent web server.

Please submit the proposal via email (a PDF file) to the course responsibles (both). By preference 10am to allow us time to read it.

The number of points depends on the ambition of the project (initial offer) and on the quality of execution (final point). Experience says, unless you mess it up you get 6 per person...

Is the bonus points awarded for the project also available on any future exam or only the first one?

The bonus points are valid for the term end exam, and only to improve a grade of E or higher. They cannot be used for improving from F to E.

"Project report: The individual project report should be at most 3 A4 pages long, and should explain your role in the project, the problems you faced and the way you solved the problems."

To be clear, does this mean we should write three different reports, one per person?

Yes, the report is individual, the presentation is 1 per project.

I can't find the assignments. Where are the assignments?

Assignments will be posted 1 week before the submission deadline.

The submission deadline is in one week. Where is the first assignment?

See above.

Thanks.

The original version of Problem 3 in HW1 had a few typos. Please consider the revised version of the homework problem set.

I just tried to turn in assignment 1 but the website says "It is not possible to add a document at the moment. Either all categories are inactive, or you already uploaded a document for the active categories". When will it be possible to upload?

I just opened it for HW1, please try again now.

It worked! Thanks.

Note to Safari users: It seems like some versions of Safari do not follow the redirect to the KTH CAS and therefore it is not possible to login to the HW system. If you use Safari and experience a problem logging in, please try to use another web browser (Firefox, Chrome, IE would work).

Hi!

There seems to have been some sort of quirk with the homework pdf: I downloaded a copy on 140904 9:53 which differs from the current copy. In problem 3 the last row in the next hop column reads 193.114.237.1 in the current version but reads 31.167.172.143 in my old copy (see this screenshot or the file).

Hi, 

Yes, as noted above, the previous version contained some typos which needed to be corrected. The revised version was posted on 140905.

Oops, missed that, sorry!

Are there any instructions on how many points to give for each problem? Right now I can only find the solutions, but not on how many points I should give. I suppose I should not give 0 points for one small mistake in problem 2.

Hi. You should configure your https certificate so it is accepted. It's a bit embarrassing for an internet course to not be able to manage this...

It is not about not being able to set up a signed certificate, it is not more difficult than to set up a signed one. When the server was first configured there was no official way to get a free signed certificate. Now there is a way, but support has been non-responsive so far and thus we have not been able to obtain one yet. We will change the certificate as soon as we get one. Hopefully within a few days.

@Dennis: For each question there is number of points allocated, but within that we do not have any specific instructions, it is left to your fair judgement.

Two questions:

1) I think a there was an error in the wording of question 2. It says you should assign subblocks to A-J and not A-K. I didn't see that until now. The problem is that in the homework I'm now correcting, the student has assigned addresses only to A-J. I should consider this correct then I guess?

2) Perhaps not that improtant, but does your solution to question 2  really fulfill the stated requirements? You have borrowed 2 blocks of 8 addresses each from network F, yes? Which means the number of host addresses for F is: (512-2) - 2*8 = 494 which is less than the minimum 500 addresses required for F?

/Erik

Unfortunately, the question was indeed formulated as A-J, thus it should be considered as correct.

It is actually kind of important, it did not. The revised version (borrowing 1 block from E and 1 from F) does.

Thanks!

I think I found another typo in the questions for Homework 2. In 1b) it is asked what happens after the datagram has been delivered to host H_1 - which makes no sense? It should read H_2?

/Erik

Definitely, thanks!

The documents aren't anonymized. Is that really okay?

I have some questions/opinions/suggestions about the grading.

I might be missing some grading information from some other source on the course site but the solution set is not enough for this case. The student I'm grading, as well as many other students I'm sure, have errors in an otherwise correct solutions and I do not know how to grade this. The system then asks for a numerical value with no possibility to indicate that the grading is in need of a second opinion and I am not comfortable with that.

I really wish the system would allow comments, or some way of indicating the need for a second opinion. It would also be appropriated if with the next homework, a guide how to handle small error would be included (ie forgot to mention X, -1 p). It could also have been more clear in what format you wanted the grading, I had already made a neat table with my grading before realizing I needed to add the same document as the one I had received. 

I really like the course and I think the homework is a really great idea, I'm just not comfortable grading someone else in a subject I am just learning myself. 

What if we find our grading from the other student wrong? Can we do something?

How do I submit gradings?

Through the homework submission system. Please check the 'help' to see how.

@Dimitrios: We will consider these cases after HW4 is graded, if you miss those points to reach the 300 limit. Otherwise it does not really make a difference...

@Matilda: I know it is not easy to assess how many points a partially correct solution should be worth. Nonetheless, I think that figuring it out in itself it is a very good way of learning. Please give it a try.

@Lucas: Do you mean that the grader is visible to the author of the solution? Only if you include your name in the grading. The other way round I do see why it would be problematic.

what happens for those who haven't got their submissions graded?

You will be able to see the result latest on Monday.

In HW2: Question 5 a) reads: "What is the MSS used in TCP?" - by "TCP" you mean "this TCP connection" described above, yes? Not something like TCP in general?

Thanks!

/Erik

It says used "by" TCP, and it refers to the connection.

About HW2 question 3, I'm not sure what is meant by a "routing STATE". Is it equivalent to a routing table, i.e. are you simply asking for a list of "destination, metric, next hop" belonging to a certain router? Maybe this is trivial but i found no such concept in the course book and I want to score my 300pts :)

Yes, that is what we mean with it, essentially the knowledge of the router at this point in time.

In question 5: "It takes 1ms for TCP to generate a segment (with or without data) and this can be done in parallel with sending a previously generated segment".

Should we assume this is true also for generating an ACK, (i.e. does the definition of segment contain ACKs)? Or should we assume an ACK is sent immediately?

Should we also assume only one segment can be generated at a time (not in parallel)?

In HWK2 question 5 about TCP, it is not mentioned if the connection is IPv4 or IPv6. Should I make assumption? The fragmentation will be different because of different header size (20 bytes vs. 40 bytes).

Yes, but your assumption should be consistent with the path MTU given in the problem...

@Emil: Yes, both of your (first) interpretations are correct.

About HW5 Q5, it says:

"The size of a segment having a TCP header only is negligible in terms of transmission time."

Do you mean that for ACK:s and the like, count only round trip time, but for actual data transmissions, also include the bytes/(BPS), i.e. time needed to stream out the bytes? If so, does this count once (while sending) or twice (sending and receiving)?

Yes, the transmission time should be accounted for.

Sorry to say, but that didn't quite clear things up. Or is it too detailed a question? I honestly don't know whether this transmission is to be counted once or twice for a packet.

You need to account for the transmission time and for the propagation time. Try to make a drawing, that will help you to figure out whether count the transmission time once or twice.

When sending a tcp packet containing data, can we assume the transmission time of the header itself is negligible?

Yes, that would be rather reasonable.

When calculating all the RTO values in task 5, should we assume that the first RTT measurement was made during the handshake process (like it's done in the book)?

Question 1 of HW2 mentions that "ARP snooping is enabled", and I was wondering what that meant. Does that mean that all hosts intercept unicast ARP packages and add the addresses to their cache or something different? The book doesn't explicitly say.

@Emil: rfc6298 does not exclude that so may assume the first measurement is done during the handshake.

http://tools.ietf.org/html/draft-ycheng-tcpm-rtosynrtt-00 has some discussion about the advantages that this may have (note that this is not an rfc, it is merely a discussion).

@Tobias: You are close, the lecture slides on ARP contain the answer. As a hint: the Ethernet network interface card (NIC) discards a frame that is sent to a unicast address other than the NIC's own address. You have to switch on promiscuous mode for the NIC to pick up all unicast frames (but that you usually do not do because it loads the CPU).

@Emil/György: Does this mean i CAN also assume that no such measurement has been done? In this case, the rfc doesn't specify default values for SRTT and RTTVAR so I guess they must be 0 (or null?)?

@Isak: The rfc does specify the initial values (not 0).

When an acknowledgement arrives that acknowledges two segments (or more) at the same time I have noticed that the lecture notes says:

CWND = 1MSS
– Send 1 segment
– Receive 1 ACK, and increase CWND to 2MSS
- Send 2 segments
– Receive ACK for 2 segments, and increase CWND to 4MSS (I assume this one ack for two segments)

and also it says "For each ACKed segment, CWND += 1 MSS"

while the textbook says
"Remember, for each ACK, the cwnd is increased by only 1 MSS. Hence, if three segments are acknowledged accummlatively, the size of the cwnd increases by only 1 MSS, not 3 MSS"

I followed the books instructions until I realized the book and the lecture notes seem to say different things?

@Mikael: (my opinion, not official truth)

I had the same worry, but my guess is that the textbook is right, and that the lecture notes saying 'receive ACK for 2 segments' should actually be 'receive an ACK for each of the 2 segments', so no accumulative ACKing.

The lecture slide does not consider delayed ACKs, and to be explicit on row 4 it should probably say "ACKs for 2 segments".

The authoritative source is rfc5681 Section 3.1: "While traditionally TCP implementations have increased cwnd by precisely SMSS bytes upon receipt of an ACK covering new data, we RECOMMEND that TCP implementations increase cwnd, per:

cwnd += min (N, SMSS) (2)

where N is the number of previously unacknowledged bytes acknowledged in the incoming ACK."

Note that this is recommendation and it protects against ACK inflation (which would be possible if following the book).

Regarding the grading of HW2, Q5:

How important are the exact time stamps? If someone writes 913, 914, 915 instead of 915, 916, 917 , because of some minor interpretation difference, is that fatal? Also, should the time for each package be as important as the SRTT etc? There are a lot of points in the balance here, and i would appreciate some refining hints for the grading of this question.

Is this reasoning wrong?

"The speed of the network is 4 Mbps, 4 000 000 bit/s = 500 000 bytes/s = 500 bytes/ms.
This is how fast we can send data. As each segment is 1000 bytes it will take 2 ms
to send one segment."

/Erik

Sorry, I saw the last comment now that the above reasoning is correct (the comment about missing the transmission time of 2 ms being a mistake). But then can you really send segments at +904 and +905, doesn't it take 2 ms to send each one? It should instead read +904 and +906, right? And then you have to redo the whole figure (well almost)... just like we had to do a couple of times, haha! :)

/Erik

It does take 2ms to send each segment. The solution was partly using 1ms for the transmission time. +904 and +906 would be correct. The revised version will be available shortly.

I followed the lecture slides and used CWND 1, 2, 4, etc. I didn't see the question here about this, but does this still count as a mistake then?

Well, it is not really correct as per the book and the rfc...and unfortunately, sometimes the book is incorrect itself. Note that you only need 300 points, take this as a learning exercise, it really is.

I'm sorry if I'm not understanding something but this is bothering me:

In your solution of the 2. task in hw2 you say there is 3508 bytes of payload to transfer via IP, but then in the solution you only transfer 3500.

Also, something completely different. In the same task, you calculate offset regarding only bytes of data in the IP datagram not including the header. In the book, it says "...relative position of the fragment with respect to the whole datagram...". And the whole datagram should include the header as well I think

You are absolutely right on the first matter.

The offset is not relative to the beginning of the IP header of the original datagram, but to the beginning of its payload. "Whole datagram" in the book refers to the "whole payload" of the original datagram.

In the solutions there is no answer about part of question 4c), namely:

Does it affect the usefulness of the ICMP timestamp for the purpose of clock synchronization as used in (b)? 

Should we just skip that part of the question or..?

Additionally, I am unsure about how to correct question 3c), would it be correct to send 2 different messages on the north interface, one to A and one to D?

What will happend to those who do not get 300p totally on the four homeworks?

@Evelina: It seems like we have a version problem between the problem set and the solution set, as the solution set does not include this last question, and hence no explixit answer either.

Since the underlying assumption (symmetric delay) does not hold, the method would not work properly. This is why we added the discussion about algorithms/protocols that would work.

Regarding RIP2: RIP2 is supposed to use multicast on BMA links, such as the one on the North interface of the router. On NBMA networks and point-to-point links RIP2 can use unicast addresses.

@Eric: There will be an extra HW assignment, which you can use to reach 300.

From the looks of it, the calculated RTOs using the handshake in the answers are completely wrong. First of all we we're supposed to calculate the first 4 segments, in the answer only the handshake and the three first segments are calculated. Secondly the third value would have to be recalculated even though the RTT is the same since the SRTT has an updated value (in 2) and is used for calculating both the new RTTVAR  and SRTT (in 3) - right? But the roundtrip times are messed up in the solution. The handshake (R'_0) should be 601 ms (this should not be included in the answer though!), then R'_1 is 903 ms, R'_2 is 603 ms, R'_3 is 603 ms and finally R'_4 is 903 ms.

Or have I misunderstood this completely?

/Erik

The solution provides the RTOs for the first four segments sent (note that 2 segments will have the same as no new ACK arrives before the second one is sent).

So I did completely misunderstand - thanks for clearing it up :)

I interpreted the question as we were supposed to give numbers for the first four segments of data, which you clearly didn't write - just to be clear.

/Erik

In homework 3 question 2e, should we assume all three objects are on the same web server?

In homework 3 question 3e,

"Suppose that the client you are using is strictly sequential, so that it fetches one object at a time."

How sequential do we have to be? Can we request the next object as soon as we have retrieved a previous one, or do we have to wait for the connection to be properly closed first?

HW 3 question 2e: All three objects are on the same server. 

HW 3 question 2e: You could include the time it takes to close down a connection, but you don't have to. There is actually a grading comment (which you haven't seen yet) that states that either way is fine, both are considered correct. 

HW3 Q4:

The book doesn't seem to say exactly what the client does when it receives an IP address from the server; it 'isn't allowed to use the address' until the ACK has been received, but does 'use' include announcing source address of messages from the client?

Also, I worry about the image for Q4. Why is everything given an explicit address except the DHCP server, and why are some hosts labeled with random letters while others aren't?

HW3 Q4: Regarding the use of the received IP address, I suggest that you take a look at the lecture notes. When it comes to the address of the DHCP server, it is given in the diagram you are supposed to fill in.

You can disregard from the random letters - the figure was borrowed from another context so the letters do not mean anything for this particular problem.

Oh, right. I was looking for it in the first figure... Also found the appropriate lecture notes :)

Hi!

I was just about to continue on homework 3 since the deadline is tomorrow and I won't have time to work on it before 17:00. The problem is that the link to the assignment is broken. I really need as much points as possible go get enough points to be allowed to do the exam and this really hurts my chances.

I've also noticed that the links are sometimes broken, especially for Macintosh users. Something about proxy error...

I've uploaded it to http://www.csc.kth.se/~emille/Homework3.pdf if you can't reach it at the moodle site. (I will remove after the deadline).

Great, thanks!

There is no broken link that we are aware of, and no known issues with accessing the web from a Mac. If you are using a proxy, that is a configuration of your browser,. It is not related to the course web.

If you have problems accessing the course web, please be specific about the problems you experience and we will try to help out. 

Hi! In HW3, what exactly is meant by configured port in exercise 3? Is it just a port that a server listens to but that is outside the range of well-known ports?

It is part of the problem to figure out that some servers need to have ports for which there are no generally agreed upon default values. Hence, as part of the work of setting up this system, those port numbers need to be assigned. Those are configured ports.  It is not about the range, the user might decide to pick an unused system port for instance.

So any service running on any port that is not a well-known service running on its well-known designated port then? :)

But "for which there are no generally agreed upon default values" DOES mean configured ports are outside the range of "well-known ports", right? For example, in HW3 Q3 can we use port 22 for the SSH forwarder in the first connection? Wouldn't it be easier to make port 22 well-known no matter it's for local communication or for global communcation?

@Elias: Well, I am not sure how I would phrase a precise definition given that terminology is not entirely well-defined. In this question, what we are after is that there is one port for which the user needs to decide the port number. There is no default for it.

@Haosheng: Yes, if you don't have a regular SSH server running on the SMTP client host, you could decide to use port 22. It would work, although I am not sure it is a very good idea. The important thing to realize is that it is something you need to configure. There is no default, well-known port for it.

Keep in mind that the forwarder port on the SSH client is not the same as the well-known port for SSH, so if a solution uses 22 for this port, without providing an explanation that clarifies that in this case it is actually a configured port, it should render deduction in points.

Hej,

the link to the solutions to HW4 has been added, but right now we are still lacking permission to access it ("Sorry, this activity is currently hidden").

Regards, Andreas

HW4 solutions are now unhidden. Sorry for the mistake.

I was wondering.. Can AH be used with a UDP header in the answer to question 2a or does it only work for TCP (since the solutions only mention TCP)?

It can be used by both, or in fact any protocol ;)

From RFC4302:

   In transport mode, AH is inserted after the IP header and before a
   next layer protocol (e.g., TCP, UDP, ICMP, etc.) or before any other
   IPsec headers that have already been inserted.  In the context of
   IPv4, this calls for placing AH after the IP header (and any options
   that it contains), but before the next layer protocol.  (Note that
   the term "transport" mode should not be misconstrued as restricting
   its use to TCP and UDP.)

Yes, I have seen that.. I just wanted to make sure I had not missed anything and that UDP would actually work too since the solutions did not include it :)

I don't want to sit on my high horses, so to speak, but I do think the information about IKE is generally a bit off... Particularly about the cookie which seems to be set ONLY by the responder and also ONLY if it has a lot of half-open connections (thus suspecting that a DOS attack might be in action), not always. Especially, there doesn't seem to exist any cookie at all on the initiator's side, as indicated in the slides and in the homework.. But if I'm wrong, I'm glad to admit it and learn about how it is :)

RFC on IKE: https://tools.ietf.org/html/rfc5996

Maybe the solutions to HW4 are somewhat condensed. Please have a look at the lecture notes from the IPsec lecture where we discuss IKE. IKE is a rather complicated protocol and in the lecture notes we tried to introduce complexity in a step-by-step fashion so there are certainly simplifications made in several of the slides.

In slide 33 we just want to introduce how a (stateless) cookie can be used and in that example we only look at the responder cookie. In the second slide about IKE phase 1 (IKE Phase 1 - Main Mode cont'd) I have tried to illustrate are more complete behavior of IKE, using of both initiator cookie and responder cookie.

I understand... but according to the RFC, it seems that's more of a fallback mode rather than the standard mode of operation... Is this how it's used or is it, even though not explicitly stated in the RFC, used more often than that in reality?

From RFC 5996:

    When a responder detects a large number of half-open IKE SAs, it
    SHOULD reply to IKE_SA_INIT requests with a response containing the 
    COOKIE notification.

There are different variants of main mode and there are also something called aggressive mode so it is difficult for me to say (off the top of my head) what is mostly use in reality. The way of sending cookies in message 1 and 2 (as illustrated in the slide) is taken from a book where one of the co-authors also co-authored RFC 5996. However, the example in the slide is older than RFC 5996 and these things develop over time. There have been several RFCs before RFC 5996, which have been obsoleted by RFC 5996.

After IKE was introduced it was found out that there were problems using stateless cookies so one explanation might be that the recommendations for how to use cookies have changed along the way.

Gotcha! Thank you! I just got a little bit confused when I read the RFC, but computer science is not straightforward :)

I have 2 questions about homework 2 exercise 5 TCP:

 1) CWND increases according to the formula cwnd += min(N, SMSS) as specified in RFC 5681. SMSS, the sender maximum segment size, is 1000 bytes in the exercise. According to RFC 5681 SMSS is the size of the largest segment that the sender can transmit. This value can be based on the maximum transmission unit of the network, the path MTU discovery [RFC1191, RFC4821] algorithm, RMSS, or other factors. In the exercise do we take into account in all those aspects or do we simply say that SMSS = MSS?

 2) The second time that 3 segments are sent: Why are there only 3 segments sent and not 4? Actual segment size = min (CWND, RWND) min (4000, 6000) = 4000 so what am I missing? (I guess that it has something to do with congestion control.)

I can't login on the submission system. Might be a problem with kth.se login since it took me several retries and in total 15 minutes to login to kth.se. I've cleaned my browser between the retries.

It probably is, because I could login without a problem. Please let me know if you continue experiencing the problem.

@Sam: 1) SMSS is the sender MSS, so we can simply set it to MSS as computed based on the MTU.

2) It has to do with cwnd += min(N, SMSS), since we use delayed ACKs, the second ACK will acknowledge N=2000 bytes, but the SMSS=1000, so you can only increase the cwnd by 1000 bytes. Delayed ACKs affect the way congestion control works.

I am still confused about question 2). I have understood the way CWND increases but what I don't understand is why the actual window size isn't increased (to 4 segments with total 4000 bytes) the second time when 3 consecutive segments are sent from A to B.

You cannot send 4000 bytes if your CWND=2000+1000=3000 bytes.

But the second time 3 segments are sent CWND = 4000 bytes (3000 + 1000). It is marked in the figure in the solution that CWND = 4000 bytes (or is the solution incorrect?).

The reason is that there are 1000 bytes outstanding at that point (not yet ACKed bcs of delayed ACKs), so even though CWND is 4000, you cannot send more than 3000 bytes.

Thank you! Now I got it. :)

Are lab sessions individual or in groups?

They are in groups of 2.

You will be able to sign up for Lab1 starting tomorrow (Wednesday).

Does both students register for the same time slot or is it enough with 1 form the group?

Yes, everyone signs up for the lab individually. You can then team up once you are there. You should also submit individual solutions to the preparation questions.

Will there be any more labsessions available? All labsessions have been full and I haven't been able to sign up for any.

There will be more sessions next week, and there are a few slots available.
Anyone who still hasn't signed up, please try to do so; you should be able to see options now and if you cannot find any, please contact me at ljubica@kth.se and I will assign you manually.

The reading list for lab 1 includes chapter 12, which hasn't been covered in lectures or suggested reading, save for section 12.2. Is that sufficient for this lab or should the entire chapter be read beforehand? Will it be on the exam?

No, please read the chapters on UDP and TCP instead. The list of chapters was from 3rd ed., will fix it now.

When can IK2218 students expect the link to be available to register for Lab2?

You will be able to sign up from tomorrow (Sept. 23).

Should we write one lab report per 2-group or one report per 4-group?

Normally a group of four writes a single report, except for the cases where only two students performed the entire lab (if there were no other groups to merge with).

Last Sunday, you said that there would be more lab sessions for the next week. Did you mean sessions for lab 1 or lab 2? I was trying to register for lab 1 but it was full. Will there be more lab sessions available for the first lab?

We need to see how many students haven't registered timely. It is likely there will be a make-up session for the first lab on October 16.

Is it possible to open up 4 slots on a lab session the next week for me and my 3 friends?

Thanks in advance, Kristofer.

How do we know if we passed lab 1: UDP and TCP? Is there a system used to relay this information? My group sent in the lab report and have not heard anything since then.

Once we have all the reports graded, we will start sending feedback to everyone.

You will be able to sign up for Labs 1 and 2 starting tomorrow (Wednesday).

When will the lab manuals for lab 2 be published? There is only 5 days left until the labs so it would be good if we could start preparing for it soon :)

Thank you for the reminder! It is now available.

To where do we send the lab reports?

The lab assistants said to me to pick one random lab assistant's email address (can be found at https://www.kth.se/social/course/EP2120/). You should probably send to an assistant that attended your lab session.

Yes, please send it to one of the lab assistants that attended your session. In case of Labs 1-2 either to Ljubica or to Ognjen.

Hi, do we have to send a report for lab 1 EP2120? Last week, we asked to one of the assistant, and he said to us that we didn't have to send a report for lab1. 

In EP2120 you do not need to send a report for Lab1.

When will the lab instructions for lab 3 be available?

Lab 3 instructions are available, Just follow the link given above for Lab 3 DNS.