Episode: 1350
Title: HPR1350: The Origin of ONICS (My Intro)
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr1350/hpr1350.mp3
Transcribed: 2025-10-17 23:59:42

---

One.
Hello, hacker public radio.
This is Gabriel Evenfire, a long time HPR listener, and I thought at about time that I submit
my own show to the community. Why? Well, I think everybody knows the answer to that. There's
all the fame, the glory, the money. Well, mostly the narcissism, I suppose in my case.
I just have this project that I've been working on on my own time at home, and I love to
gush about it. I talk about it all the time to my family, to my friends, to my co-workers,
and apparently I'm willing to babble on about it to random strangers in the internet. Hopefully,
if the little project isn't of interest to you, at least the tale of how it came about
might be. If not, well, there's always tomorrow on HPR. You never know what you'll see next.
My project is called ONIX, ONICES, which stands for Open Network Inspection Command Suite.
The idea behind the project is to create a suite of tools that one runs on the command
line to manipulate network packets, in the same way that your traditional UNIX tools,
like said and awk and grep and TR and so forth, manipulate lines of text. There have been
various command line networking tools over the years. There was a suite called SPAC, way
way back 10 years ago for generating packets on the command line. There's the venerable netcat.
There is HPing, but none of these I thought really provided the flexibility and power that
we tend to see again in the usual UNIX command suite. So I thought I'd give it a shot.
Once upon a time and a older job about nine years ago, I did a attempt to create a few of
these tools and I was trying at the time just to create some framework for us to create test
cases for a larger project that we were working on. And when I was done with them, our customer said,
hey, you know, it would be a really good idea. If you guys open sourced those tools because they're
probably not worth money, but you know, it would get us a lot of attention in the community,
people would probably like it. And the company I was working for said, open what now? Why would
we give away our software for free? Oh, I was less than pleased about that. And I thought for a
while to try to get them to see the light and they never did. A few years later, I left the
company. I have no idea what became of that code base. And quite frankly, it was done in all the
wrong ways. But it still gave me the hope that creating a tool suite as I was envisioning was
at least possible. So what's the idea? How would it work? The general idea is that you would start
with set of files like PCAP files, traces of network packets. And you would pass them through these tools.
And you would manipulate them. You would get to fold spindle, mangle, and mutilate them. You could
extract them. You could split them into separate files. I was pretty much hoping that I could find or
create an equivalent tool for each of the following programs. Cat, Grip, Said, Ock, Sort, Diff, Head,
Tail, WC, Word Count, TR, T, and Paste. And of course, I'm sure there are many others that came to mind at
the time too. Unfortunately, as I was looking at how to create these replacement tools, I kept getting
hung up on one important thing, the PCAP format. Quite frankly, it's terrible for this kind of an
application. Every PCAP file has a fixed set of fields. It's not very extensible. Every PCAP file has
a file header on the front that defines attributes for the entire packet trace. So for example,
you can only have one data link type in a packet trace. So if you have packets that were captured
some from Ethernet and some from PPP, you're out of luck. They can't go in the same PCAP file.
Furthermore, the fact that you have that header in front of all of the packets means you can't
just cat two of these files together because then you have an extra header stuck in the middle
that the tools can't parse. It also makes it a pain to split because every time you want to
say, take a packet out of one file and say, shove it into another. You have to remember, oh,
by the way, what was that stupid header? Now I have to slap that in front of it.
So the more I fought with this poor file format, the more I realized I just had to replace it,
at least for my tools. So I created a packet capture format or an external representation for
packets that I call XPacket. So the first two utilities that I ended up creating were utilities
to take in PCAP files and translate them to and from the XPacket format.
I thought this was a great idea because since the PCAP library is able to read and write packets
from the network itself, then these utilities would also be able to read XPackets right off the
network and write XPackets right to the network. After I had these programs done, the next thing I
started working on were some libraries to actually parse protocol headers and I had a goal in mind.
I wanted to not have to rewrite the protocol parsing for every one of these tools. So I was trying
to create a set of libraries that would let one interact with the fields and the packet headers
in a network packet without actually having to know how they were put together so that
those same parsing and manipulation routines could be called in every single tool I created.
So I came up with a library based around the abstraction that each header would be parsed to a set
of offsets into the packet buffer. The idea would be that you could have offsets that would just
be marked invalid for certain fields that might be optional within the packet and if you had variable
length fields within a packet, you could use two offsets to denote their start and end.
Why do this? It seems extra complicated when all you could just do is say it starts here and you
give it a length dynamically as you parse the packets. Well, one of the things I definitely wanted to
be able to do in the tools was insert or remove bytes from the middle of the packet.
And whenever one did this, one would have to adjust all of the starting offsets and possibly the
lengths in every single packet parse. So by instead modeling the parsed information as a set of
offsets, the insertion and removal routines actually could just shift those up or down as needed
without having to know what they referred to. After I wrote that library, I also wrote another
library that would allow one to refer to the fields within a protocol by name and this library would
provide all of the information on how to extract or how to write to these different fields,
just given the name of the field that you were looking for. Now, any tool, if provided a name
by on the command line, would know exactly how to extract that field from the packet and how to
embed that field back into the packet if you wanted to change it. Once I'd written that, the next
part were some basic packet buffering libraries, which appear in just about every networking project
in the world. And then I was ready now to create utilities to dump out packets that I had in
X packet format so that it was in a nice human readable form. In other words, I wanted to be able to
do TCP dump. So the next thing I decided was that instead of having this little TCP dump utility
look like traditional TCP dump, I was going to have it output the packets in an annotated
hex dump format that was similar to hex dumps that are used in traditional Unix command line tools.
That is, there would be comments, comment lines that would contain information about the packet
fields, like they would tell you that a set of bytes referred to IP address 1.2.3.4.
And then below that would be the actual hex dump of the packet that contained that data.
The idea would be that one could, if you wanted, go and manipulate the hex by hand with your
favorite text editor if you wanted to change the packets. And one could use traditional hex
manipulation tools to convert that hex data right back into packets if you wanted.
So you could bring regular Unix tools to bear onto the packets in some small way.
So I wrote the routine to dump the packet into that hex format. And then another one,
just for good measure, that would dump from the hex format back into the packet format.
Great. That worked. That was all well and good.
The next thing I wanted to do was to be able to multiplex and demultiplex packets.
I wanted to be able to, ultimately, if I wanted to, write a router out of some of these tools.
So that would mean that if a packet came into the system and I had a script that did a route look
up and it said it had to go out this interface instead of, you know, eth0, for example,
instead of eth1, that there would be a tool that along the way would say, ah, okay, to get to
eth0, I follow this path and to get to eth1, I follow this path. And again, I wanted this on the
command line. So what one would be doing is multiplexing and demultiplexing on named pipes in the system.
So that's where the next two tools came along, packet mocks and packet demux.
And now it was time for our favorite, everybody's favorite Unix tools, the ones that you
learned to love and hate at the same time, but once you know them, love, love, and love. And that's
your regular expression matching tools, like GREP or SED, or for that matter AWK.
So I wanted to be able to match patterns on packets and then carry out some sort of actions
when those patterns were matched. At first, my approach to this was to create a little stack-based
virtual machine. The idea was to be similar to the Berkeley packet filter virtual machine. I
wanted it to be completely verifiable so that one could write a program in it and it would be so
safe that you could even embed it inside the kernel. But I wanted it to be able to do more than
just match. And so as I was working on the instruction set, I kept fact the control store was
non-rightable, just like the BPF engine, but I ended up with something that looked nothing like BPF.
I ended up with a virtual machine that had a segmented memory model, a non-rightable control
store, a verifiable instruction set for a subset of the instructions so that the virtual machine
could run in a verifiable and non-verifiable mode. I also added features that is instructions
in the VM that are a bit higher level, like the ability to pull bytes out of the packet based
on indexes into the packet parses that my other parsing libraries would generate. Also, I had
added a regular expression engine for matching in the packets, which was another side project from
years before. It's my own reg X engine. It's not very sophisticated, but it does work. And also
the ability to do bulk copies in and out of packets and bulk string comparisons and mask string
comparisons. Okay, so now I had a virtual machine that the virtual machine itself didn't understand
anything about TCP or IP, but it received a library that could parse TCP and IP and so and IPv6
and ICMP and so forth. And if the program running under it knew how to ask it in the right way,
it could nevertheless extract fields from the TCP header or the IP header or the IPv6 header
or modify fields in said headers. So the virtual machine itself knew nothing about the protocol
parsing, but the programs could. And at this point, as I was testing this virtual machine, I'd
basically been hand coding all of the assembler in my test programs. And then one of my co-workers,
I saw had written this quick and dirty assembler and I thought, well, if he can write a quick and
dirty assembler, why in the heck can't I? I can just do this and knock it out in a weekend. So,
what did I do? Well, we were visiting my in-laws that weekend and any moment where we weren't
socializing or having fun, I went back to my laptop and I tried to grind out a quick and dirty
assembler. And then while I was at it, I also added a disassembler so I could verify that the
assembler was doing what it was supposed to be doing. And then after that, I realized, well,
assembling and disassembling is nice and all, but I also need some sort of a file format similar to
ELF or something so that I can actually save the assembled instructions into a into a format that
can be read in and executed. And I need a runtime that has the net VM in it to be able to run on
packets. Okay, so it took a little bit more effort than I thought, but ultimately, within a week,
I had an assembler or runtime file format and an interpreter that could read that file format,
pass packets through it and run these little net VM programs on every packet.
And so I thought, great, cool, now I can manipulate packets in a programmable way in the sort of
form that I had wanted with set an awk where you could have, you know, a pattern that was matched
and then based on that pattern, it could carry out operations or it could just run code on every
single packet. And just like awk also has begin and end segments, you could also just write code that
ran at the beginning of the program before packets came in or after the program finished after all
the packets had been read in. But even as I finished this off and was padding myself on the back,
it occurred to me that this really wasn't as high level as said or awk. I mean, come on,
who is going to write assembly on the command line? Okay, so I really needed a higher level
language and I decided that it was high time to create it. So I settled on creating the language
PML. I call it PML for packet manipulation language or more colorfully, packet mangling language.
Again, the idea was that this language would be an awk like language. So it would be higher level,
it would be touring complete. But the structure of the program would look like a pattern action
format. Originally, I used a Lex and Yak to write the lexical analyzer and the parser generator.
But I really, really didn't like Yak that much because its parsers were single threaded on most
platforms and just rub me the wrong way. And then while I was listening to the BSD talk podcast,
I heard about the lemon parser generator. And the lemon parser generator, you probably haven't
heard of, but it's actually a parser generator embedded in every distribution of SQL Lite because
it actually compiles the compiler for the SQL language in SQL Lite. And this parser generator
was released public domain. So one is perfectly free to take it and embed it in any project
regardless of the license. I thought this was pretty cool. So I ripped out my Yak grammar and I
replaced it with a lemon grammar. And success, well, I hadn't really finished the grammar yet,
but I was getting somewhere. The grammar was coming along, it was parsing, and I had the feeling at
this point I'm almost done. Once I get this awkward language, that's sort of the crown jewel.
Everything else will just fall out from that. As we'll see, I was half right.
So now I was really grinding away at this tool. I was working late at night after the kids and
my wife were asleep. I was coding on weekends. I was coding during my daughter's basketball practice.
I was just really hammering on this because it seemed like the end was in sight and I really,
really wanted to build this tool. My family in this time was becoming less and less pleased with my
obsession. But finally the parser was done and then I started on the code generation and I thought
that would be simple and boy was I wrong about that. The code generation took as long as the parser
and the lexer. But I did grind through it and understand that the idea was I was compiling
programs to the net VM virtual machine. When I had my first hello world in PML working,
that was really fun. I was really happy about that. But what really got me to do an engineer's
victory dance was that only a few hours later I managed to get a program compiled that would
go through and compute the internet checksum on an IP packet and verify whether it was correct or
not. So that involves loops, it involves in direction in the program as well as printing and
control flow and all of that was working. So I was doing something right. So I was really pleased.
PML was coming along and I realized as I was doing this however that there were a lot of cases to
cover and at this point I decided I had to do something to manage the complexity and I did what
I really should have been doing all along and that was to start writing unit tests. Lots and lots
of unit tests, unit tests for the lexical analyzer, unit tests for the parser generator, unit tests
that would check to see whenever the net VM instructions that the compiler generated differed
from a previous version in the exact same program. Then of course I started testing to make sure
that the programs actually worked and did what they were intending to do and for each one of these
I built a regression test suite. The good news for me was now that every time I added a feature I could
just type make in the test directory and it would run a whole battery of tests and make sure that
everything worked exactly the way it had before and it gave me a lot more confidence. The most
programmers don't really think or think about or like doing testing. I through this process just
came to love tests because it made my life so much better in the long run at least for this project.
Okay so I kept grinding away at this. PML was functional. I thought now is the time to start
open sourcing this. So I made an account on Gatorius for it that it was about this time that I
decided on the name Onyx. I thought there are so many good programming languages that are named
after gems like Ruby or Pearl. So Onyx seemed like a nice one. Plus since it was a Unixi command suite
I thought Onyx sounded like a nice Unixi name. I started showing this off to my friends and co-workers
and unfortunately this was a reality check for me because the more I showed it off to them the more
they thought it was pretty cool but I realized just how rough it was around the edges.
So I started grinding through and doing some more tests and some more tests and I started
working in particular on an entire section of the code that I had written but never tested and
that was packet generation. You see I wanted to be able to create a set of PML programs or at
least scripts that ran PML that would let you build up a packet a little at a time. The idea
would be that you could cat some data through and pipe it to a RAP TCP program that would put a
TCP header around the data and then pass that to a RAP IP program which would put an IP header
around that and then a RAP Ethernet program which would put an Ethernet header around that and each
one of those programs would let you manipulate the fields to customize them as you needed. So for
example like setting the Ethernet MAC address or setting the IP time to live or whatever.
And then finally you could even pipe it after that you I mean you could redirect it straight
out to a file if you wanted to save your work or you could pipe it right to the PCAP out program
to send it on the network so you could even generate a packet onto the network straight from the
command line. As I started creating these tools and these scripts which are really just thin
wrappers around PML I found lots and lots of bugs in that untested portion of the code and a bit more
grinding later and a more unit test later and now they were done and now I was thinking okay now
really I'm about done right. At this point I even went one step further and I created a little
script called TCPcess for TCP session and this script would let you give a list of files
on the command line and it would generate a complete TCP connection starting with the
connection establishment with the SIN, SINAC and ACK packets followed by the data in the files
that were given on the command line broken up into appropriately sized TCP segments
and responded to by the receiver with acknowledgement packets until all of the data was sent in both
directions and the command line flags would specify what order those files were to be played in
each direction and then finally finishing with a shutdown handshake of the FIN ACK, FIN ACK
nature and I thought that was really cool it was another engineering victory dance when I ran
this program and then I piped the output of this program to my pcap out file to turn it into a
pcap file and then I piped that into TCP dump to read it back and TCP dump looked at it and said oh
yeah this looks like a perfectly well-formed TCP stream that was pretty cool oh so okay good
now I really had some flexible tools they could do some really neat things and I thought surely
I'm I'm almost done but I realized that I still had some things that I wanted to take care of
I am obviously as you can tell a bit of it do it yourself or when it comes to programming
I like to have written everything myself it's just one of my hangups if I can help it I want to
do it myself and so I didn't like the fact that my programs depended on the Lexiop library and the
lidpcap library I didn't mind so much about lemon because lemon I could distribute with the code
itself so while I didn't write it myself and I have some aspirations in the future of writing
my own parser generator for now it was okay at least if somebody wanted to download and compile
my tools they didn't need to download the parser generator but they did have to have Lex and they
did have to have a lidpcap and I didn't like that I wanted the whole thing self-contained I hate
autoconf I hate doing dot slash configure I want the thing to be able to compile right on the
command line the first time in every environment that supports C if at all possible so what I did was
I wrote my own pcap parsing library to be able to parse pcap files or write pcap files and then
I went and I basically rewrote the lecture by hand as a C program well more of C library okay good
so now the system could be self-contained right just to make sure that I could capture packets off
the network live and inject them onto the network and not depend on libpcap I then took the extra
step of going and finding the Linux and BSD ways of capturing and injecting packets and then I created
versions of the pcap in pcap out program that would capture packets from the network or send packets
to the network but I created a separate version for each of those OSs and then the build would select
which one to compile based on some big detection of your environment and in any case the pcap specific
programs and these OS specific programs were optional builds and so if the build system couldn't
figure out whether you had libpcap and if it couldn't figure out that you were trying to compile for
Linux it would just create dummy versions of those programs that would error out immediately and say
sorry that's not supported on this platform you could still run all of the other tools in the
tool suite so again I was trying to make it as platform independent as possible by isolating
the platform specific parts from the rest of the code base and now I was thinking great great
now I really should be done and you know what all this time in the back of my mind I've been
thinking there is one killer app that does not exist anywhere in the networking world that I've
ever seen and that's the equivalent of diff for packets if you don't know the diff program takes two
files and it finds the minimum set of transformations to convert one file into the other file so it shows
you how to edit the first file to produce the second file in other words it shows you the
difference between the two files and I thought and I knew that this would be really really cool
to have for packets back when we had done the older version of this of this suite back in nine years
ago somebody had come along and taken the tools we had and they wrote their own diff utility
and it was pretty cool it was neat it relied on however the packets having an embedded sequence
number so that the program could tell when packets had been dropped or when they had been inserted
so I thought well I could rewrite something like that so I wrote a bunch of scripts that would
attempt to embed sequence numbers into a packet stream in various spare bits and fields within the
packet and then extract them at some point later but the more I thought about it the real
I the more I thought this is kluji and really finding the edit distance is a simple programming problem
at least it is when you study it in school but it's a little trickier with packets because
modifications to the packet stream can happen at least three layers of abstraction entire packets
can be dropped or inserted or reordered that's one layer also headers can be pushed on or popped
off for example when a packet enters a tunnel or leaves a tunnel and third individual fields
within a packet header can be inserted or removed and of course individual fields within a packet
can also be changed for an example of fields that get inserted or removed from a packet consider
IP options or IPv6 extension headers so in other words to find the minimum amount of editing
that was needed to change one stream of packets into another I had to figure out whether it was
cheaper to drop a packet outright or change a few of its fields to make it match a different packet
in the stream and this all seemed very daunting so I kind of stalled and put it off for a while
and on the side I decided you know what let's take a break let's go and learn javascript
I know many programming languages that one was not in my repertoire I'm partial to lots of
scripting languages but I'm not much of a web programmer myself so javascript just never I didn't
have a good reason to to learn it so I learned javascript and while I was doing it I thought you
know what why don't I dust off my memory and my algorithm to textbook and pull out our typical
domero levinstein edit distance algorithm and code that up in javascript and I did and I realized
you know what this algorithm isn't really that tricky come on let's just buckle down and do
diff for packets so I did I started in earnest and I had a little bit of a time trying to engineer
the the cost functions to be able to determine again whether it was cheaper to make one type of
edit versus another but I did settle on one that I thought was sensible that the cost would be
proportional to the number of bits inserted or removed relative to the size of the packet
and with that and with these multiple layers of running the same domero levinstein distance
algorithm I had a really cool diff program I could put in a stream of packets and I could pass
it through a pml program and the pml program might say okay drop the fifth packet duplicate the
eighth packet and on the tenth packet change the ttl to five and when I ran that I would you
know I had an infile it would go through my little script and then I would have my output file
and then I'd run both files through pdiff and I would see exactly that it would say oh here
in drop the fifth packet and look here insert this packet after the eighth packet which looked
exactly like the the eighth packet oh and by the way in this what's now the eleventh packet
the ttl was changed to five so cool this was really great and I was thinking this this is something
I don't even see anywhere else in the networking world this is going to be awesome really I should be
done now this is this is cool and so as I'm wrapping up and documenting different pieces of my
code in in particular documenting the pml language I was thinking about hey by the way how would I
write ping in this tool suite and as I thought about that it occurred to me that the pml language
and another my tools really had any sense of of time-based events they were all run to completion
that is a packet comes in you do stuff on it until you're done and then it goes out
it's time me for a little bit but then I thought hey I have actions in pml that occur when the
program starts in the begin pattern and then I have bits of code in the pml program that runs
when the packets when all the packets are finished going through the system that's in the end block
of the pml program so why don't I just create a new type of block a tick block and this will execute
code every so often let's say once a millisecond and then as long as your events have to happen no
at no finer granularity than a millisecond you can have any sort of time-based events and this
actually turned out to be a relatively easy mod so I put it in and great now I had a really cool
new feature and at this point I was thinking I really was thinking this is you know all that I
wanted to get done for my basic functionality I should be wrapping up now and about this time
it was time for our family vacation to the beach so I was ready to relax and chill and although
in previous summers I had been working on these tool suites I thought I'm probably not going
to be doing much this time but my daughter she had been my eldest daughter to be specific she had
been saving up her money to buy her own first laptop and she wanted to get a Google Chromebook
and she did and I was very curious about this little box so I was wondering what sort of
network traffic I'd see from it in the background so while we are at the beach house I'm on the
wireless network and I break out my tools and I'm scanning the packets looking to see you you know
who's who's chattering with whom and how often and I realized you know it was getting a little tedious
to be looking at these packets on a packet by packet basis it would be nice if I had a program
that would look at the stream of packets and just create a summary of the connections that we're
going through you know oh look this TCP connection started oh and now it's ended oh this TCP
connection started oh it's still going and it's transferred this much data oh it's still going
and it's transferred this much data oh and now it's done things like that and also wouldn't it be
cool if I could tag the packets that we're going through with information about which connection
they were on so that other tools could operate on that specific information I had a nice
extensible metadata format to do that so why not do it okay well you can guess what happened
late nights at the beach house spare moments between lunch and and heading down to the pool
whatever it was I worked on yet another tool called this one NFT for network flow tracker
and within a day and a half I had it done and working exactly the way I want I was pretty happy
about that and I started looking again at the traffic and I was noticing some cool interesting
patterns I was pretty happy about this and then as I'm watching the packets I'm realizing oh
you know it's always one more thing my ICMP and ICMP V6 parsing really were very very basic and
wouldn't it be nice if instead of having to decode these packets by hand it actually went the
extra mile and decoded them based on their type in code appropriately okay so another day or so
was spent going through the RFCs looking at all the relevant type codes and making sure I was
parsing them all correctly and augmenting my protocol parsers to handle that okay and that worked
and now I was thinking this is great everything is solid and polished and I can't think of any real
other must-haves that I have to add for this tool suite awesome so really I should be done I'm done
now right the next week I'm in the office and I'm talking with one of my co-workers and we
were talking about markup languages for our documentation at work and he was talking about you know
even enroph is a nice serviceable language and I thought to myself yeah it is you know I had to
do my thesis in t-roph because it gave me great control over the text layout and let me
fulfill the university's text layout requirements and then as I thought about this I thought you
know it really wasn't that hard to write that I really should have man pages for all of these tools
sigh all right fine before I go on bragging more I I'm going to create man pages for all of them
okay so I broke down this was about labor day weekend this year and I decided to start grinding
through the weekend and getting all the man pages written at this point I had about 28 tools and
scripts 10 10 scripts 18 tools so I I ground through it and I got all the man pages and I really wanted
these to be useful man pages so I put examples and all of them and I tested to make sure the examples
worked and as probably many of you have had this experience as I started to document it I realized
that there were some more rough corners in the tools and their options and and things of that
nature and so I went back and and I ended up giving a lot of the tools a little bit of a not an
overhaul just a makeover to again stand out some more rough edges so it turned out I really wasn't
done even then finally I said okay now it's time to do a podcast on these tools and I sat down and
outlined this podcast and after I was done I don't know what made me think of it but I realized
there was another key command line tool for text that I had no equivalent for in Onyx and that was
sort I couldn't sort packets I just didn't have any capability for it so all right I will knock out
sort I already have sort routines written from my younger days and so all I have to do is
figure out what I'm sorting on and I'm done so I wrote a P sort utility for packet sort
and at the time it would only sort on packet metadata but that was okay because the metadata
was extensible I thought you know you could if you really wanted to just extract fields from
the packet put it in the metadata and sort on that and so I thought that's good enough I'm done
and then the next day I thought now I should at least have a bash script that automated that process
so I wrote up a quick bash script that would take a set of fields that you wanted to sort on
and embed it in the metadata and pass it to the sort routine and then delete it from the metadata
so it wasn't lingering around and that led to polishing up some of the metadata manipulation
in the virtual machine in the net VM and also in the PML language and then I realized you know what
that was still clunky the P sort program itself really needed to be able to understand the keys
or rather take the keys directly and then I realized even that was clunky really the P sort
utility itself should be able to directly take the fields that you wanted to sort at and build the key
so okay I went through I figured out a way to using those same protocol parsing libraries extract
the key from command line arguments and you know what it worked it worked exactly as I had hoped
it didn't actually take more than a day to get this utility up and running
then I deleted the extra script that I had as it was very superfluous at this point
so now I had P sort great I should be done now really done now well there was I will admit one
little bit further of hackery that I had to perform before I finished this podcast and that was
I had never really done a podcast and I needed to figure out how to make the audio quality for
this decent and by the way I hope it is decent but I can tell you one thing regardless
of how it sounds to you it was a hundred times worse when I made the first recordings
terrible terrible I tried different mics that I had one on a headset one that is the mic
on my laptop itself one on my tablet and then one with this old gaming mic that I had
that's on an arm and all of them sounded pretty bad so I thought well let me just use the mic
that seems to be the least objectionable that is this this little gaming boom mic and at least
maybe audacity can make up for the rest perhaps I can just filter out the background noise and so
I was having problems with especially spikes that would clip whenever I accidentally breathed
into the mic so I tried holding the mic off to the side and talking with it that way and
it was better but then my voice was very soft and it really wasn't great so my last bit of
hackery was I got a piece of Terry cloth from our drawer full of rags I got some electrical wire
some industrial electrical wire that I'd been using for outdoor wiring which is nice and stiff
I got the screw top to some jars that my wife uses to jar vegetables for Christmas gifts and so
forth and what I did was this is you know the screw top of course has a hole in the middle so I
wrapped the Terry cloth around that fastened it with rubber bands and then attached that using this
nice stiff electrical wire to the mic itself so I thought this would keep my breath from making
these spikes and then I could speak close to the mic and the noise reduction would be better
and like I said it worked I'll put a picture of this in the show notes it's quite a little ugly
job but it definitely helped I hope this audio quality isn't too objectionable to you but like I
said whatever it sounds like to you it really was much worse so I was kind of proud of that little
hack job as well all right so I've come to the end of this long description here and am I really
really done with this project have I got it in a state where there's not too much to add
well I have to confess no I thought about adding random number generation to the PML language I
thought about adding backing store for the packets I thought about extending the language with the
associative arrays I thought about adding DNS resolution into the tool suite I thought about
ways that I could improve the connection tracking I have plans to improve the packet ordering
or reordering detection in PDF I have ways already in mind for how to improve the performance
I have visions of embedding my little net VM inside the Linux kernel or on an embedded platform
and say my Raspberry Pi that I got for my birthday I have more protocol parsing that I can add
I can improve the multiplexing and demultiplexing I can improve the string formatting
I just know that this project is is not going to end anytime soon and I think my wife and
kids will be less pleased with this I think at the end of the day I've had a lot of fun with this
project anyways it's a genuine hacking project it was I created it to scratch and itch it has
led me on all sorts of tangents that I never imagined in the first place it has definitely led me
on some emotional roller coasters sometimes feeling like I'm the most brilliant person in the world
and sometimes feeling like this is just idiotic and dumb and nobody's going to want this
it's kind of taken on a life of its own but at the end of the day I look at this project like I
look at all hacking projects and that's through the lens of the advice of my advisor who said
when you're picking a project to work on just pick something that you yourself would want to use
and think would be really cool because if you build it then you've already won if other people
like it then that's just icing on the cake so I can definitely say that this project is in a state
where I'm using it and I'm having fun with it so if you out there are interested in playing around
with it you can find onyx at getorious.org slash onyx oh and I see us there's a wiki page in there
that gives you some information for how to get started I'll just give you right now the quick
version of how you build it and how you install first you type get clone get at getorious.org
colon catlib slash catlib.get and this will clone my sea libraries that I've been hacking on
for longer than this onyx project those libraries are fundamental to the use fundamentals of the
operation of the onyx programs themselves the next step clones the onyx repository get space clone
space get at getorious.org colon onyx slash onyx.get so now you'll have two new directories
in the directory that you perform these commands in one will be onyx and one will be catlib.
first cd space catlib slash src then type make and that builds the catlib then cd space dot dot slash
dot dot slash onyx then run make and this will build all of the onyx programs and it will also run
all of the regression tests if test 42 43 something like that fails don't worry about it that one
is timing based and I've had a devil of a time making it always always work just run make again
and it'll probably pass if you're really brave and you or if you find the tools intensely useful
you can also then say type pseudo space make space install to install the program in user local
bin and user local man and then you can have it in your path and use it to your heart's content
I hope you found this story interesting I hope the audio quality comes out decent and in general
thank you to the hbr community for years of enjoyable content I hope this is received as enjoyable
content as well and all of you guys out there keep on hacking this is gabriel even fire if you
would like to reach me you can reach me through the guitarias dot org site for onyx or you can email
me at even fire eve and f i r e at sdf dot org thanks a lot guys bye
you have been listening to hiker public radio or hiker public radio does our we are a community
podcast network that releases shows every weekday on day three friday today's show like all our
shows was contributed by a hbr listener like yourself if you ever consider recording a podcast
then visit our website to find out how easy it really is hiker public radio was founded by the
digital dog pound and the infonomicum computer club hbr is funded by the binary revolution
at binwreff.com all binwreff projects are proudly sponsored by luna pages from shared hosting
to custom private clouds go to luna pages.com for all your hosting needs unless otherwise stasis
today's show is released under a creative commons attribution share a line lead us our lives