In 1998, I proposed push-based search indexing to Microsoft. They rejected it.
In 1998 I was 20, living in a small Russian town called Velikiye Luki, connected to the internet via dial-up. I had a Pentium 60 and saw a problem: search engines were wasting enormous resources crawling static sites that rarely changed.
I didn't know the term "push" and had no access to academic papers — I learned later that Xerox had formalized push concepts back in 1987, and Netscape had server push in 1995. But in 1998, the web search world was 100% pull-based. So I built a prototype that applied the idea to indexing: a Windows filter driver intercepted IRP_MJ_WRITE calls, a lightweight parser extracted text and links on the host, and only the compressed delta was sent directly to the search engine — eliminating crawling for static sites.
I sent a physical letter to Microsoft. In October 1998, a reply arrived from Redmond. Five lines:
"Although your submission is interesting, we have determined your product does not fit our current business needs."
I didn't stop. I found two private investors, got $1,000, named the company Grabit. Then I went to the US embassy for a visa. I brought printouts of correspondence with my partner. One phrase killed it: "If everything goes well, he will stay in the States." Visa denied. Grabit died before it was born.
I wrote to Yandex. They thought I was building a data theft tool. I met a provider in a St. Petersburg chess club. He nodded and said they were "planning something similar." Nothing happened.
Twenty years later, the industry arrived at the same approach for search indexing. Google launched PubSubHubbub. Microsoft and Yandex launched IndexNow in 2021. Cloudflare built it into their platform.
I didn't earn a cent. I didn't become famous. But I kept the letter.
Sometimes the approach isn't wrong. The timing is.
Full story with photos of the original letter and technical details:
