Archive for May, 2006

This animation nicely sums up how it feels to deal with Charter, my cable modem, and being forced to rely on dialup Internet access on a daily basis.

Thought I’d post a screenshot of what a typical day’s connectivity looks like in my household this week. This is a shot of the window of my connection monitor program. The top box shows the dates and times at which the every-minute ping batch moved in and out of what it considers to be a “good” connection.

The bottom box shows a graph of the connection’s quality. The Y axis runs from 0 pings lost per minute at the bottom to 10 pings lost (100% loss) at the top. Each pixel on the X axis represents five minutes of test time.

Mousing over the graph shows a tooltip indicating the time and pings lost tally for the sample at the corresponding time. Since each pixel represents five samples, the value for pings lost is actually the average of five ping batches.

Today I spent some time playing with (or trying to) the new stuff in JavaScript 1.7, having finally gotten my hands on a build of the JS shell based on the 1.7 alpha branch. Didn’t have tons of luck; I don’t think all the new stuff has actually landed on the branch yet. Hopefully soon.

Having been stymied on that front, I decided to spend the rest of my day pawing through the NeedsEditorialReview category on MDC. Got three or four entries revised before coming up on some stuff I need to research more fully.

In particular, I haven’t spent any time in the past looking at the plugin API. At a quick glance (the most I felt like giving it at midnight!), it looks like there are two APIs, the older of which is deprecated. But I’m not sure, since the docs that are currently in place seem to merge the two together in what would be quite the unholy union were that the case.

Which of course leads me to suspect that my first quick glance is misguided. I’ll have to look into that some more.

I’ve sort of taken it on as my mission to clean up and clear out everything that’s currently tagged as needing editorial review, whether by fixing what needs fixing or removing articles that are unneeded (or obsolete).

People are starting to go, “Oh yeah, we have a guy around to do this now,” and are now asking me to do editorial passes or rewrites of material. That’s fantastic! Makes me feel like I’m earning my keep. Keep the requests coming!

For folks that might notice me futzing around with text they wrote — sometimes I’ll make really minor editorial changes because I think they’ll improve the use flow of text. Don’t take it personally. I’m picky about grammar, style, and the quirks of punctuation. Fortunately, that’s part of my job. Hopefully we can make our documentation really shine, both in terms of thoroughness and accuracy as well as readability.

This post blabs a lot about last night’s episode of Lost, so if you don’t want to know, skip it. If you do want to know, highlight the rest of the article to be able to see it.

So the “?” on the map is apparently the Pearl Station. With nine video monitors for observing other facilities. That’s quite interesting. Looks like only one station’s monitor still works, which is the Swan, where the survivors are located.

Probably the most intriguing thing, however, is the confluence of the dreams shared by Eko and Locke. These two have apparently been “chosen,” for lack of a better word, to accomplish something together. It will be interesting to see what comes of that.

So this morning, after being offline more or less all night, Charter calls and asks how things look, and while I’m on the phone with them, I realize that the reason for the errors I was seeing was a configuration change done yesterday while testing.

Once I undid that change, I was suddenly getting phenomenal performance!

For about an hour.

Now we’re back to 90%+ packet loss and I’m having to use dialup again in order to get any work done.

There’s unfortunately nothing I can do about it today, as there’ve been a couple of general outages in our area, and since that might be the problem today, Charter pretty much doesn’t want to talk to us. I can more or less understand that.

But if it’s still not working in the morning, I’m going to be furious.

My baby girl just took her first steps! Go, Sophie, go!!

I’m continuing to run my app that monitors my cable modem connection, under the assumption that it’s going to keep sucking.

It’s based on a Perl script Dave Miller wrote ages ago, back when we were having similar problems with Syndicomm’s ISP. I turned it into a nice little Cocoa app for my Mac. Objective-C is fun.

It works by sending out ten ICMP echo packets (pings) every 30 seconds. If at least three of them come back within two seconds of being sent, the connection is considered to be online. If the connection goes offline (ie, less than three packets come back), it takes seven packets coming back in one batch for it to be considered online again.

It then records all this information, spitting out the ping log into one window and maintaining a log of the date and time of each online/offline switch. Online events are recorded in green, offline events in red. It also maintains a total time elapsed counter as well as indicators showing the total number of pings sent, received, and lost, as well as a percent packet drop overall.

With a little tweaking, it might be a handy utility to share with others, so I might do that. Needs the stop button to actually work, and the box that lets you enter an address to ping currently doesn’t do anything — it always pings a hardcoded site right now.

At the moment, the connection has been on since around 6:30 PM. Unfortunately, that’s about an hour after the cable company left today, so they’ve fairly obviously not fixed things yet.

On top of that, although currently packet loss is low (around zero, most of the time), packet round-trip time is very long. Earlier today, they were averaging about 3.5 seconds, with the longest round-trip time being almost 30 seconds.

My current ping stats look like this:

64 packets transmitted, 58 packets received, 9% packet loss
round-trip min/avg/max/stddev = 427.334/1204.596/18346.937/2309.430 ms

That’s just not pretty at all. On the other hand, even though it’s incredibly sluggish, at least I’m able to move data around at the moment. I don’t expect it to last though.

I’m mostly finished with my first rough-cut draft of a document outlining the new features in JavaScript 1.7. Once I get responses to a couple of emailed questions, I hope to wrap it up and send copies to select people so they can point out that it’s all wrong and that I need to rewrite most of it.

This tends to happen a lot when documenting brand new technologies. Existing proposal and spec documents tend to be out of date by the time the technology is actually implemented, leaving the technical writer in the position of making educated guesses and then letting the engineers point out how it’s all wrong.

All in a day’s work.

Fortunately, in this case, the guys know good and well that the specs and proposals are out of date and are ready for this. That makes life easier.

Still duking it out with the cable company over my cable connection. My little app indicated several outages overnight, but at the moment, it’s been online since around 6:30 AM. However, packet loss is starting to climb again and is now at around 50%, so I expect the connection to fail sometime in the next 30-60 minutes.

Good times.

These guys are incredibly incompetent at maintaining their network.  It spent most of the day offline today, then came back on at around 6:30 PM, only go go off again at 8:30 for about an hour and a half.

My connection watcher needs tweaking still — it senses the occasional dropped packet and treats it the same as a real outage.  I need to change it to only consider it an outage if the drop rate spikes, even if not every packet is dropped.

Tomorrow I’ll go somewhere that has more reliable Internet access to work, and let Sarah fight with Charter over this thing.  Probably my mother in law’s house again.

This has now been going on for three weeks.

In a final desperate act of “let’s get some hard evidence to show Charter,” I spent a couple hours today writing and fine-tuning a Cocoa app for my Mac that pings out onto the ‘net every few seconds and tracks the times at which pings start and stop being returned. This way, I can map out when my connection fails and then when it comes back online, down to a fairly small margin of error.

It’s been running for 45 minutes now and has currently 8% packet loss. It hasn’t actually cut off completely yet; that’s when I’ll get the most useful output out of it. Right now, my log shows up and down every few seconds or minutes, which doesn’t help much.

Time to get some sleep. Here’s hoping it doesn’t crash during the night…