Many years ago I worked for a small software development company. Let's call them InfoStuff, which is not their real name. We were bought by a larger group and the responsibility to construct a group WAN fell on myself and a colleague. We were programmers who had to program during the day and learn all about networking in the evening. Somehow we managed to set up a countrywide WAN and after a while it seemed logical to start a small Internet service provider. It was targeted at businesses using leased lines but a small number of dial-up accounts were available for certain staff members and naturally I had one of these.

The ISP side of InfoStuff grew, acquired more staff, grew some more, and so on until the group sold the ISP business along with the InfoStuff brand name. Being a part of the development business, my colleague and I stayed with the group. The group continued to use the newly sold business as our ISP and I still had my free dial-up account, as did one or two other of our staff members. Some years passed and our group moved to a new ISP, but still my free dial-up account kept working. I kept expecting it to be cut off or to start being billed for it but I never was. More time passed and the availability of the service gradually fell off. It seemed that the number I dialled into at the local POP was not one of their standard dial-up numbers, having a different login ID format. Sometimes their dial-up server would hang and not be reset for days. I obviously couldn't complain and didn't really mind as I had the number of another POP 1500km away and as I only am on a telephone scheme that limits after hours call costs to a fixed maximum, it cost me the same to phone this POP. A few months ago the local POP stopped answering at all and I started calling the distant one exclusively. Then on Friday evening I was connected and the line dropped. When I tried to reconnect all I got was Line busy. The dial-up server was hanging. By Sunday morning I was suffering from Internet withdrawal and as I needed to do some online banking I came in to the office to get on the net. At the same time I started looking for another ISP.

I asked for recommendations on the mailing list of my local LUG (Linux User Group) as I wanted to avoid M-WEB, South Africa's largest ISP, who are notoriously Windows-centric. I got several suggestions and also looked at <a href=www.polka.co.za<, who have advertised quite a bit on TV, and Webstorm who came up in a Google text ad when I searched. One of the LUG members suggested his company, Frogfoot, a smaller one running all Linux equipment. Although they were a little more expensive than the others the idea of supporting them appealed to me. Sadly their web site did not have an online sign up and there was no answer on their 0860 number. I was not keen on Polka as they are just another face of the M-WEB Borg so I looked at Webstorm, who happened to also be the cheapest at R69 per month. I started their online sign-up procedure and gave them all my name and address details. This page was not secure but I wasn't too worried until I got to the next page where I was to enter my banking details. This page was also not secure and there was no way I was going to send my banking details unsecured so I gave up on them. Lastly I had a look at Webmail ISP, suggested by another LUG member. They were a few rand more at R79 per month but offer more, giving 4 mail boxes instead of the single mail box and aliases offered by others. I signed up with them and was able to connect when I got home. I was connected until late last night and the speed seemed to be a little slower than I used to get with InfoStuff but that might have been just the connection I had then as dial-up speeds are always quite variable from connection to connection.

I am quite comfortable with the idea of using bit masks to test variables for specific values. What came up the other day is the need to test a byte for the presence of only a single set bit, where the specific bit is not known.

I wanted to know whether the byte had any of the following values, without caring what the actual value was;

    00000001
    00000010
    00000100
    00001000
    00010000
    00100000
    01000000

The consensus from everyone I've discussed this with is that a loop with multiple tests is required, but I can't help thinking that out there somewhere is someone that has a clever bit mask or mathematical way to test for this with a single expression.