Comments inline. > On Jul 14, 2020, at 3:35 PM, Andrey Khomyakov <[email protected]> > wrote: > > I was once asked at a FANG interview how I would affect incoming traffic > using BGP. I listed the usual offenders like AS path and med. He kept asking > how else, to which after pondering I said that I cant think of other ways > right now. He was insisting I find one, so I theorized on using more specific > prefix even though that?s technically not a BGP method.
Using the Origin comes to mind, but according to RFC 4271, that shouldn’t be done. IMO, it’s not fair to ask candidates to do things that the generally accepted reference material strongly cautions against. > I think that was when he decided I failed, and I decided that he wanted me to > name a creative method he himself used at his megascaler. Considering I never > had to even think about solving problems at their scale I think that was just > dumb to insist I come up with it. > I?ll never know what that magical method is since he didn?t actually give me > the answer I didn?t know ?\_(?)_/? I would not have been happy if an interviewer didn’t tell me the answer [s]he expected. At the very least, I would have asked if I could refer to RFC 4271 during the interview. > The interview in general seemed to be more about technical nerd knobs of > random technologies and less about how I think about problems. > > My take is that nerd knobs can be taught and/or looked up. How a person > thinks and dissects the problems at hand can hardly be changed easily. I agree that how people think about problems is something that should be more important on interviews than specific answers. That said, in this particular case, the interviewer could have been looking for a response such as “the tie-breaking procedures in BGP route selection are not limited to what is documented in RFC 4271.” There are people who have implemented such procedures, some of whom may be on this list. (I have seen NANOG listed as a interest of theirs on their LinkedIn profiles.) Perhaps, if they are reading this thread, they would comment on what type of response they are looking for, if they ask such questions. > The question about ?what happens when you enter google.com into your browser? > is probably my favorite and I ask it to candidates every time. That and ?how > can you confirm a remote system is listening on a given TCP port?. Idk what > it is that confuses people, but close to zero people answered ?telnet or > netcat to that port? and most of them went down the ?I ssh into the system? > or ?I look in the firewall? or ?I netstat?. I’ve never asked anyone that question, but it’s been asked of me a few times. Before saying anything else, I ask what specifically is the interviewer looking for — how google.com <http://google.com/> is resolved by DNS? how the HTTP request (if HTTP is the method) is handled? how packets move from the machine the user is typing on through the Internet? There are “entire universes” of discussion that could be had in response to this question. I’ve also been told that it is highly recommended (at some companies) to ask questions in response to questions that are open-ended, because the companies don’t want to hear “stackoverflow responses” that could have been memorized. > Another one is ?what is the summary prefix for 10.0.0.0/24 and 10.0.1.0/24?. > So.many.minds.blown! Some ask for pen and paper. Others just start doing > binary conversions out loud.... I don’t necessarily think this is a bad thing. The candidates may be being careful because they don’t want to make a mistake on an “easy” question, fearing that getting it wrong could drop them from consideration. They also might regularly use tools that do this sort of thing, and don’t regularly have to answer that type of question without these tools, in much the same way that people (generally) use calculators instead of pen and paper for “simple” arithmetic. > Turns out there are as many bad candidates as there are interviewers. I > consider myself a bad interviewer so I try to shy away from interviewing > candidates if I can. And based on this email chains so far it seems like > there is not an agreed upon good interview method. Perhaps there will never be an agreed upon good interview method. I wish this was communicated to the mainstream press and politicians who have been led to believe that companies can’t find the right people because of insufficient education. This may be true, but it is not guaranteed to be true. —gregbo
