Hacker News new | past | comments | ask | show | jobs | submit login

"Deprecating" might be a better term than "forgetting." Google's business isn't driven by long tail content. Probably it never was.

Maybe somewhere there is a Google disk with the the hash of an exact phrase the author typed into the search box. But statistically, that hash won't be found in hot memory vector space when cosine similarity runs on a nearby server. Finding the phrase would require a batch job that runs much longer than the engineered time limit Google imposes on search queries. Without a "let me know in 24 hours" option, Google's search will partition data into what should and shouldn't be accessible. That partition will always be according to Google's business goals. All the information may be indexed, but only the fraction of the index beneficial to Google will ever be accessible to ordinary users.

The crux of the story is that there is no business case for Google to return the author's web pages in search results even if the Wayback Machine implies that Google could.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: