When she was alive, she did everything to help her children, even to her own detriment - just like my father. She deserves a better destiny than vanishing into eternal night, into nothingness. Just like my father does, who died thirteen years ago.
So now I can see at least one method of reversing death within the limitations of present-day physics.
Two assumptions must hold true for that method to work.
The first one is a requirement for "unbounded but finite" amount of computational power. Something like Kardashev Type II or even Type III civilizations will be able to reach.
We know, as a fact, that information about past state of our Universe is accessible to us in the form of Cosmic Microwave Background radiation. This information essentially tells about initial distribution of mass and energy from which modern galaxies and stars evolved and eventually Earth and humans.
There is also 21-cm hydrogen radiation and neutrino backgrounds that give the same information from even earlier epoch.
From that information we can build an increasingly more precise model of earlier Universe, and then run a simulation to see how it will evolve. Another, virtual Universe, of sorts. With enough computational power and accuracy in simulating physical laws, we can simulate it to any level of detail and see galaxies, stars and planets form, together with any lifeforms evolving on those planets.
Now, this simulated Universe is not guaranteed to evolve in the same way as our own one. Stochastic or chaotic behavior in simulation rules may drive it to some entirely different state after all. But with "unbounded, but finite" amount of computational power we can run more than a single such simulation.
Furthermore, we can check the discrepancy between any simulated Universe and our own using the (partial) information we know about past states of real Universe (like expected distribution of galaxies, mass and other properties of simulated Earth vs the real Earth, genomes of virtual lifeforms vs the real one etc). Such information fragments that are identical between the real and virtual Universes can be called "checksums" or "hashes" and can be used to select among the multitude of simulated Universes the version that is closest to our real one.
These "checksums" can include known information about the humans as well. Their memoirs, their known biography details, their writings, photos, videos and other data. Everything we know about their lives. The selection goal, ultimately, is to have a "lagging" timeline of our own Universe as a copy in virtual form.
Now, this is a pretty big "if" to state those assumptions will hold true. Access to "unbounded, but finite" amount of computation power seems like the easier one, at the very least, humanity must persist and expand beyond the home planet, eventually colonizing Solar System and reaching nearby stars. But the required amount of calculations for precise virtual Universe simulations may demand to use supermassive black holes or even artificial quasars as energy sources. This may take many thousands or millions of years to reach.
As for the second assumption, there is no obvious solution to confirm that humans in virtual Universe will have the same mind states as their originals in real ones. The best "checkpoint" data available to us are the cryogenically preserved bodies and brains in cryonics labs. If all those brains are ever scanned to atomic level and their internal structure is found completely identical to their virtual copies in simulated Universe, we can say with high confidence that any other brain in virtual Universe is the same as their original in real Universe.
Implementing all of this project is not a trivial task and may take millions or even billions of years. Required technology may even be not that hardest part, remembering and caring for long gone loved ones for all that time might be the harder challenge. But so what? If death itself can be defeated, would not this be the greatest achievement of humankind ever?
Among the endless night, a ray of light now shines for me.
P.S. Some notes on computation requirements for such a simulation . We probably do not need to simulate entire Universe at the same level of details. Causally unconnected events in the remote galaxy a billion light years away didn't have much effect on people in Ancient Rome. So we only need to simulate those areas of Universe at high enough resolution to replicate the observable events, such as Crab Nebula supernova.
For simulating Earth, the full resolution is required only for the thin shell centered on Earth surface, say +/-15 kilometers from geoid average. Nobody really is affected by how exactly Earth mantle is simulated as long as accurate plate tectonics is achieved and known volcanoes erupt on the right dates. Also simulating other planets in Solar System as single points of light would be sufficient until telescopes appeared. The dark side of Moon doesn't need to have any features until 1959, etc, etc.