Paul Graham Debriefing: Viaweb June 1998

Paul Graham Debriefing: Viaweb June 1998
Hours before Yahoo sold out in June 1998, I took a screenshot of Viaweb. I thought it would be interesting to look at it one day.

The first thing you'll immediately notice is how compact the pages are. In 1998, the screens were noticeably smaller than the current ones. If I remember correctly, our main page was placed just in the standard window that most users opened at that time.

Browsers back then (IE 6 didn't arrive until 3 years later) only had a few fonts, and those didn't have anti-aliasing. If you wanted the page to look good, you needed to process the displayed text into images.

You may have noticed some similarities between Viaweb's logos and Y Combinator. When we started Y Combinator, this was our internal joke. Given how simple the red circle is, I was surprised how few companies use it as their logo, but later I realized why:

Paul Graham Debriefing: Viaweb June 1998

On pagededicated to our company, you may find a mysterious individual named John McArtem. Robert Morris (also known as "Rtm") was so out of the public eye after his "Worm”, that he did not want his name to be present on the site. I was able to convince him to compromise: we used his biography and changed his name. After that, he is a little calmed down on this account.

Trevor graduated from university around the same time that the Yahoo sale took place. Thus, in 4 days he managed to turn from an insolvent university graduate into a millionaire Ph.D. It is the article in which this event was celebrated, and became the climax of my career as a journalist. In it, I also included a drawing of Trevor that I made during that meeting.

Paul Graham Debriefing: Viaweb June 1998
(Trevor also appeared as “Trevino Bagwell” in the web designers category of our website. There were people that entrepreneurs could hire to develop online stores for them. We implemented it in case some of the competitors want to intimidate our web designers. By the way, our assumption that his logo could scare away our customers turned out to be wrong.)

In the 90s, to attract virtual visitors, you had to shine in newspapers and magazines - there were no ways to be found on the web that are now. So we gave $16,000 monthly to one PR firmto be mentioned in the press. Fortunately, newspapermen we were loved.

In our article on getting traffic from search engines (I don’t think that the term “SEO” existed at that time) we named only 7 search engines that were significant for this function: “Yahoo”, “AltaVista”, “Excite”, “WebCrawler”, “InfoSeek”, “Lycos”, and HotBot. Doesn't it feel like something is missing? "Google" appeared in September of the same year.

Our site supported the possibility of Internet transactions using the service “Cybercash”, because if we did not have this opportunity, we would have serious problems with the ability to compete in the service market. But the service was so terrible, and the orders coming in from the stores so small, that it would have been easier if entrepreneurs had switched to a telephone ordering system. We even had a page on our website that had encourage sellers to use this particular method with customerswho buy a physical product, not software.

The whole site was made as a bridge that immediately sent people to “Test Drive". This opportunity was new for us to try out our software online. In order not to show competitors how our code works, we placed CGI-bins in our dynamic addresses.

We've had several regulars. It is worth noting that "Frederick's of Hollywood" received the most traffic. We put a tax of $300/month on our largest hosting stores because it was a little financially unsettling to have users with high volumes of traffic. I once calculated how much it costs for us to provide traffic for Frederick's of Hollywood and came up with something like $300/month.

Considering that we kept all the stores on our servers (in total they gained about 10 million visits per month), we consumed, as it turned out at that time, a lot of traffic. We had 2 lines of the T1s type (throughput ~ 3Mb / s), because at that time there was no AWS. Even the nearby servers seemed too risky to us, given that something is always going wrong with them. In general, our servers were located in our offices. More specifically, in Trevor's office. He didn't want to share his office with people, so he had to share his office with six humming tower servers. We even named his office "Banka" because of the amount of heat these hulks gave off. For the most part though, his stack of window air conditioners did the trick.

For description pages, we used a template language called RTML. It had to be deciphered somehow, but actually I named it so in honor of Rtm'a. RTML was Common Lisp, which was supplemented with macros and libraries, as well as a structure constructor, which made it feel like it had a structure, an order.

We were constantly updating the software, so it didn't really have versions, but the press of the time was used to having them, so we made them up. If we wanted to become really popular, we rolled out a version numbered an integer (integer). The inscription "Version 4.0" was generated by our random number generator. By the way, the entire Viaweb site was created by our online software, because we wanted to see with our own eyes how and what the client would use.

At the end of 1997, we released a multi-purpose shopping search engine called “shopfind". At that time, it was quite complex and technically advanced: it had a "spider" that could "visit" almost any online store and find the right product.

Translation: Ivan Denisyuk

PS

Source: habr.com

Add a comment