Expensive Analytics Software WTF



  • Here's an email exchange I just received, edited to protect person/company names and to put email thread in order:

    @Co-worker To Product Support said:

    My AWS server name is 47 characters.

    It appears to be 1 character too long to fit in the server name field in the data connection menu.

    @Product Support Response said:

    Hello Co-worker,

    Thank you for contacting Product User Support. The first possible workaround to this issue is to use an IP address for the server in place of the server name in the data connection.

    Another possible workaround is to setup a new name to that IP address in your hosts file. However, please be advised that sharing the data connection or publishing it in this second workaround will not be possible unless those system have had the same hosts file update.

    A third approach is to use SQL Server Aliases. A quick search brought up the following link on this process (http://msdn.microsoft.com/en-us/library/ms175176.aspx). When you introduce the connection details, be sure to verify they match with what is set in the protocols configuration of the target server.

    Please let me know if the solutions above do not resolve the issue.

    Thank you.

    Kind regards,

    Product Support

    @Co-worker to me said:

    So, we can’t connect to data on the server directly from Product- unless we figure out a workaround.

    Any ideas on the best approach to solving this?

    Product support suggested these options:
    1. use an IP address for the server in place of the server name in the data connection
    2. setup a new name to that IP address in your hosts file
    3. setup SQL Server Aliases

    Not sure if #1 will work for AWS

    I tried #2 and it didn’t work

    I’m assuming I’d need direct access to the server for #3

    So here's some WTFs:

    1) A product we pay a lot of money for has a retarded bug where the field to enter the server name is only 46 characters long

    2) Their support staff seems to think this is an immutable fact of nature instead of a retarded bug they should fix post-haste

    3) Since my co-worker already tried the Hosts file method and it didn't work, it's likely that this product bypasses the Hosts file when doing DNS lookups! (Note: it's also possible my co-worker just typoed the IP address. But he's a pretty sharp guy.)

    Anyway, I'm adding a DNS alias for this group, which should come out to about 30 characters to "resolve" the issue. What a pain in the ass.



  • Don't you know that no string can be longer th



  • My full comment can be found in your hosts file.



  •  Who in the nine billion names of God has a 47-character name for a server? [i]That's[/i] the Real WTF. They could have used a guid and been shorter.



  •  @D-Coder said:

     Who in the nine billion names of God has a 47-character name for a server? That's the Real WTF. They could have used a guid and been shorter.

    Self-documenting DNS.

     



  • @D-Coder said:

     Who in the nine billion names of God has a 47-character name for a server? That's the Real WTF. They could have used a guid and been shorter.

    Amazon AWS dual-stack (meaning both IPv4 and IPv6) load balancer name: dualstack.lb-xxxxxxxxxx.us-east-1.elb.amazonaws.com, 51 characters.

    AWS Elastic IP DNS name, western region (what we're using): ec2-xx-xx-xxx-x.us-west-1.compute.amazonaws.com, 47 characters

    Also, this is a SQL Server connection, so add 5 or 6 characters in case you have to specify the port as well.

    <FONT size=1>-anonymized - galgorah</FONT>



  • @D-Coder said:

     Who in the nine billion names of God has a 47-character name for a server? That's the Real WTF. They could have used a guid and been shorter.

    What, God needs nine billion names but one measly little server can't have 47 characters? Did He use them all up? (That said, my mind does boggle a bit at a server name that long, but it's Blakeyrat. If there weren't some rationalized (or even rational) reason for it, I would have thought he'd wtf that at us too.)


  • Discourse touched me in a no-no place

    @D-Coder said:

     Who in the nine billion names of God has a 47-character name for a server? That's the Real WTF. They could have used a guid and been shorter.

    No, the real WTF is an arbitrary limit below the 255 characters allowed for a hostname.



  • Any idea what language the program was written in? If it was C, just be thankful they included a if( strlen() ) rather than letting it buffer overflow because they can't be bothered to look up how to handle strings properly. Boggles the mind that string handling examples still use fixed-size buffers with no warnings about overruns. At least allow a decent number of characters in that case though...



  • @kilroo said:

    @D-Coder said:

     Who in the nine billion names of God has a 47-character name for a server? That's the Real WTF. They could have used a guid and been shorter.

    What, God needs nine billion names but one measly little server can't have 47 characters? Did He use them all up? (That said, my mind does boggle a bit at a server name that long, but it's Blakeyrat. If there weren't some rationalized (or even rational) reason for it, I would have thought he'd wtf that at us too.)

    According to God (according to Clarke), you only need nine characters for all the names.

    Of course this may be Unicode or some awful thing, but still that's only 18 bytes.

     


  • ♿ (Parody)

    @blakeyrat said:

    Amazon AWS dual-stack (meaning both IPv4 and IPv6) load balancer name: dualstack.lb-xxxxxxxxxx.us-east-1.elb.amazonaws.com, 51 characters.

    Yeah, I think the TLA was throwing (many of) us. Not knowing better, I had guessed that "AWS" was the package that you were paying a lot of money for, or something. Also, probably mislead by the similarity to the name, "awstats." I know I was imagining the server name being the actual name of the server (i.e., inside your firewall, etc), not a fully qualified server name, complete with enterprisey ID like information.


  • ♿ (Parody)

    @blakeyrat said:

    2) Their support staff seems to think this is an immutable fact of nature instead of a retarded bug they should fix post-haste

    Isn't that the job of product support? Give you a workaround so that you can continue working. Now, it does seem like a silly thing, though more like braindead design than a true bug. And not fixing it is probably a WTF, unless the bigger WTF is all the stuff that relies on the name being a major PITA / risk to change.



  • @boomzilla said:

    @blakeyrat said:
    Amazon AWS dual-stack (meaning both IPv4 and IPv6) load balancer name: dualstack.lb-xxxxxxxxxx.us-east-1.elb.amazonaws.com, 51 characters.

    Yeah, I think the TLA was throwing (many of) us. Not knowing better, I had guessed that "AWS" was the package that you were paying a lot of money for, or something. Also, probably mislead by the similarity to the name, "awstats." I know I was imagining the server name being the actual name of the server (i.e., inside your firewall, etc), not a fully qualified server name, complete with enterprisey ID like information.

    If AWS reported a stat of how many times a particular resource was hit, would it be called "awshits"?



  • @boomzilla said:

    Yeah, I think the TLA was throwing (many of) us.

    Mea culpa.

    @boomzilla said:

    Isn't that the job of product support? Give you a workaround so that you can continue working. Now, it does seem like a silly thing, though more like braindead design than a true bug. And not fixing it is probably a WTF, unless the bigger WTF is all the stuff that relies on the name being a major PITA / risk to change.

    They could have thrown-in a sentence in like, "this is a defect in our product, and has been reported to the development team". I mean, throw us a bone at least.

    Edit: to resolve this thread, it turns out the Hosts file alias does not work, but using the IP directly works.



  • @D-Coder said:

    Of course this may be Unicode or some awful thing, but still that's only 18 bytes.

    Sorry, it takes 36 (unless you're into some variable coding scheme in which case it can take between 9 and 54 or so).



  • @blakeyrat said:

    So here's some WTFs:
    1) A product we pay a lot of money for has a retarded bug where the field to enter the server name is only 46 characters long
    2) Their support staff seems to think this is an immutable fact of nature instead of a retarded bug they should fix post-haste
    3) Since my co-worker already tried the Hosts file method and it didn't work, it's likely that this product bypasses the Hosts file when doing DNS lookups! (Note: it's also possible my co-worker just typoed the IP address. But he's a pretty sharp guy.)

    Anyway, I'm adding a DNS alias for this group, which should come out to about 30 characters to "resolve" the issue. What a pain in the ass.

     

    Try telling a company that their DHCP implementation is broken, that you have contacted the author of the DHCP RFC and he agrees that it's broken, and have said company tell you bullshit, it's not broken and insist they're compliant with the RFC.  Had it happen when I was still a wee network engineer.  Lost my faith in vendors then and there.  Cynical did I become, yes.

     


  • Garbage Person

    @nonpartisan said:

     

    Try telling a company that their DHCP implementation is broken, that you have contacted the author of the DHCP RFC and he agrees that it's broken, and have said company tell you bullshit, it's not broken and insist they're compliant with the RFC.  Had it happen when I was still a wee network engineer.  Lost my faith in vendors then and there.  Cynical did I become, yes.

    Which vendor and what was the issue?

     

    Seriously, people, if you're going to throw around anecdotes like this, name names and give details.



  • @Weng said:

    @nonpartisan said:

     

    Try telling a company that their DHCP implementation is broken, that you have contacted the author of the DHCP RFC and he agrees that it's broken, and have said company tell you bullshit, it's not broken and insist they're compliant with the RFC.  Had it happen when I was still a wee network engineer.  Lost my faith in vendors then and there.  Cynical did I become, yes.

    Which vendor and what was the issue?

     

    Seriously, people, if you're going to throw around anecdotes like this, name names and give details.

    APC (American Power Conversion -- power infrastructure equipment (UPSes, transfer switches, etc.)). Upon loss of link, a device should re-DHCP for an IP address. Their devices don't, so if the card ends up on the wrong VLAN it either takes a visit to the closet or loss of management until the DHCP lease reaches its half life (up to a day -- our leases are 48 hours). You can't just drop link on the switch port, change the VLAN, and bring it back up to have it get a new address. In addition, another WTF is that they ask for a host name in the configuration but don't supply it when doing DHCP. That is not in violation of spec, as supplying a host name is optional in DHCP. However, if you've got a place for it, and someone configures it, it seems only logical to supply it when doing DHCP.

    I opened a case. After going round with them for a couple of times, just for shits and giggles, I went to see if I could track down the author of the DHCP spec. I contacted him and explained what was going on and what the behavior was of APC's card. He said yes, upon loss of link the card should no longer consider the IP information valid and should re-DHCP for a new address. It can make a suggestion and request to use the old address, but it should still re-DHCP.

    I referred to section 3.2 of RFC 2131 (Client-server interaction - reusing a previously allocated network address). APC's response: "The network interface is not restarting on a warmstart. Therefore, the card will maintain its IP settings, unless the lease has expired. " Their response made no sense to me. On a warmstart, its OS is rebooting, but they don't consider that a restart of the network interface. I e-mailed Ralph Droms and he said that's not the way they should be doing it. Unfortunately, in the intervening time since this incident we migrated from GroupWise to Exchange and this part of my archive got corrupted, so I have the headers but I don't have the text of the messages between him and me. But it was my first moral victory while on the networking team. All this occurred circa August 2007.


Log in to reply